diff --git a/Cargo.lock b/Cargo.lock index 3beaf38ea..1117b5711 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -4908,7 +4908,7 @@ dependencies = [ [[package]] name = "goose" -version = "1.44.0" +version = "1.45.0" dependencies = [ "agent-client-protocol", "agent-client-protocol-http", @@ -5044,7 +5044,7 @@ dependencies = [ [[package]] name = "goose-acp-macros" -version = "1.44.0" +version = "1.45.0" dependencies = [ "quote", "syn 2.0.118", @@ -5052,7 +5052,7 @@ dependencies = [ [[package]] name = "goose-cli" -version = "1.44.0" +version = "1.45.0" dependencies = [ "anstream", "anyhow", @@ -5156,7 +5156,7 @@ dependencies = [ [[package]] name = "goose-mcp" -version = "1.44.0" +version = "1.45.0" dependencies = [ "anyhow", "base64 0.22.1", @@ -5277,7 +5277,7 @@ dependencies = [ [[package]] name = "goose-test" -version = "1.44.0" +version = "1.45.0" dependencies = [ "clap", "serde_json", @@ -5285,7 +5285,7 @@ dependencies = [ [[package]] name = "goose-test-support" -version = "1.44.0" +version = "1.45.0" dependencies = [ "axum", "env-lock", diff --git a/Cargo.toml b/Cargo.toml index 350945ea7..d506f9640 100644 --- a/Cargo.toml +++ b/Cargo.toml @@ -8,7 +8,7 @@ resolver = "2" [workspace.package] edition = "2021" -version = "1.44.0" +version = "1.45.0" rust-version = "1.94.1" authors = ["AAIF "] license = "Apache-2.0" diff --git a/crates/goose-provider-types/src/canonical/data/canonical_models.json b/crates/goose-provider-types/src/canonical/data/canonical_models.json index 7f5e02742..f0730a89b 100644 --- a/crates/goose-provider-types/src/canonical/data/canonical_models.json +++ b/crates/goose-provider-types/src/canonical/data/canonical_models.json @@ -5491,11 +5491,11 @@ "id": "abliteration-ai/abliterated-model", "name": "Abliterated Model", "attachment": true, - "reasoning": false, + "reasoning": true, "tool_call": true, "temperature": true, "release_date": "2026-01-06", - "last_updated": "2026-01-06", + "last_updated": "2026-07-28", "modalities": { "input": [ "text", @@ -5515,6 +5515,33 @@ "output": 8192 } }, + { + "id": "abliteration-ai/abliterated-model-large", + "name": "Abliterated Model Large", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-25", + "last_updated": "2026-07-28", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 5.0, + "output": 5.0 + }, + "limit": { + "context": 1000000, + "output": 999990 + } + }, { "id": "ai-router/gpt-5.4", "name": "GPT-5.4", @@ -5678,6 +5705,241 @@ "output": 128000 } }, + { + "id": "aiand/deepseek-ai/deepseek-v4-flash", + "name": "DeepSeek V4 Flash", + "family": "deepseek-flash", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-05", + "release_date": "2026-04-24", + "last_updated": "2026-04-24", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.15, + "output": 0.25 + }, + "limit": { + "context": 1048576, + "output": 384000 + } + }, + { + "id": "aiand/deepseek-ai/deepseek-v4-pro", + "name": "DeepSeek V4 Pro", + "family": "deepseek-thinking", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-05", + "release_date": "2026-04-24", + "last_updated": "2026-04-24", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 1.0, + "output": 2.5 + }, + "limit": { + "context": 1048576, + "output": 384000 + } + }, + { + "id": "aiand/google/gemma-4-31b-it", + "name": "Gemma 4 31B IT", + "family": "gemma", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-04-02", + "last_updated": "2026-04-02", + "modalities": { + "input": [ + "text", + "image", + "video", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.2, + "output": 0.5 + }, + "limit": { + "context": 262144, + "output": 32768 + } + }, + { + "id": "aiand/moonshotai/kimi-k2.7-code", + "name": "Kimi K2.7 Code", + "family": "kimi-k2", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2025-01", + "release_date": "2026-06-12", + "last_updated": "2026-06-12", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.75, + "output": 3.5 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, + { + "id": "aiand/moonshotai/kimi-k3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 3.0, + "output": 12.5, + "cache_read": 0.5 + }, + "limit": { + "context": 1048576, + "output": 131072 + } + }, + { + "id": "aiand/openai/gpt-oss-120b", + "name": "GPT OSS 120B", + "family": "gpt-oss", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-08-05", + "last_updated": "2025-08-05", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.15, + "output": 0.6 + }, + "limit": { + "context": 131072, + "output": 32768 + } + }, + { + "id": "aiand/qwen/qwen3.6-27b", + "name": "Qwen3.6 27B", + "family": "qwen", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-04-22", + "last_updated": "2026-04-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 262144, + "output": 65536 + } + }, + { + "id": "aiand/zai-org/glm-5.2", + "name": "GLM-5.2", + "family": "glm", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-06-13", + "last_updated": "2026-06-13", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 1.0, + "output": 4.0 + }, + "limit": { + "context": 1048576, + "output": 131072 + } + }, { "id": "aihubmix/alicloud-deepseek-v4-flash", "name": "DeepSeek V4 Flash (Alibaba Cloud)", @@ -7406,6 +7668,37 @@ "output": 32768 } }, + { + "id": "aihubmix/kimi-k3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-27", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 3.0, + "output": 15.0, + "cache_read": 0.3 + }, + "limit": { + "context": 1048576, + "output": 131072 + } + }, { "id": "aihubmix/minimax-m2.7", "name": "MiniMax M2.7", @@ -7567,7 +7860,7 @@ "id": "aihubmix/qwen3.7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -10139,7 +10432,7 @@ "id": "alibaba-cn/qwen3.7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -10149,7 +10442,8 @@ "modalities": { "input": [ "text", - "image" + "image", + "video" ], "output": [ "text" @@ -10711,7 +11005,7 @@ "id": "alibaba-coding-plan-cn/qwen3.7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -10721,7 +11015,8 @@ "modalities": { "input": [ "text", - "image" + "image", + "video" ], "output": [ "text" @@ -11086,7 +11381,7 @@ "id": "alibaba-coding-plan/qwen3.7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -11096,7 +11391,8 @@ "modalities": { "input": [ "text", - "image" + "image", + "video" ], "output": [ "text" @@ -11141,7 +11437,7 @@ }, "limit": { "context": 196608, - "output": 24576 + "output": 32768 } }, { @@ -11196,7 +11492,8 @@ "cost": { "input": 0.0, "output": 0.0, - "cache_read": 0.0 + "cache_read": 0.0, + "cache_write": 0.0 }, "limit": { "context": 1000000, @@ -11226,7 +11523,8 @@ "cost": { "input": 0.0, "output": 0.0, - "cache_read": 0.0 + "cache_read": 0.0, + "cache_write": 0.0 }, "limit": { "context": 1000000, @@ -11323,6 +11621,89 @@ "output": 131072 } }, + { + "id": "alibaba-token-plan-cn/happyhorse-1.1-i2v", + "name": "HappyHorse 1.1 Image-to-Video", + "attachment": true, + "reasoning": false, + "tool_call": false, + "temperature": true, + "release_date": "2026-07-17", + "last_updated": "2026-07-17", + "modalities": { + "input": [ + "image", + "text" + ], + "output": [ + "video" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 0, + "output": 0 + } + }, + { + "id": "alibaba-token-plan-cn/happyhorse-1.1-r2v", + "name": "HappyHorse 1.1 Reference-to-Video", + "attachment": true, + "reasoning": false, + "tool_call": false, + "temperature": true, + "release_date": "2026-07-17", + "last_updated": "2026-07-17", + "modalities": { + "input": [ + "image", + "text" + ], + "output": [ + "video" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 0, + "output": 0 + } + }, + { + "id": "alibaba-token-plan-cn/happyhorse-1.1-t2v", + "name": "HappyHorse 1.1 Text-to-Video", + "attachment": false, + "reasoning": false, + "tool_call": false, + "temperature": true, + "release_date": "2026-07-17", + "last_updated": "2026-07-17", + "modalities": { + "input": [ + "text" + ], + "output": [ + "video" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 0, + "output": 0 + } + }, { "id": "alibaba-token-plan-cn/kimi-k2.5", "name": "Kimi K2.5", @@ -11353,7 +11734,7 @@ }, "limit": { "context": 262144, - "output": 32768 + "output": 98304 } }, { @@ -11386,7 +11767,7 @@ }, "limit": { "context": 262144, - "output": 16384 + "output": 262144 } }, { @@ -11514,7 +11895,7 @@ "id": "alibaba-token-plan-cn/qwen3.6-plus", "name": "Qwen3.6 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -11570,14 +11951,14 @@ }, "limit": { "context": 1000000, - "output": 65536 + "output": 131072 } }, { "id": "alibaba-token-plan-cn/qwen3.7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -11587,7 +11968,8 @@ "modalities": { "input": [ "text", - "image" + "image", + "video" ], "output": [ "text" @@ -11602,7 +11984,7 @@ }, "limit": { "context": 1000000, - "output": 64000 + "output": 65536 } }, { @@ -11718,7 +12100,7 @@ }, "limit": { "context": 196608, - "output": 24576 + "output": 32768 } }, { @@ -11773,7 +12155,8 @@ "cost": { "input": 0.0, "output": 0.0, - "cache_read": 0.0 + "cache_read": 0.0, + "cache_write": 0.0 }, "limit": { "context": 1000000, @@ -11803,7 +12186,8 @@ "cost": { "input": 0.0, "output": 0.0, - "cache_read": 0.0 + "cache_read": 0.0, + "cache_write": 0.0 }, "limit": { "context": 1000000, @@ -11900,6 +12284,89 @@ "output": 131072 } }, + { + "id": "alibaba-token-plan/happyhorse-1.1-i2v", + "name": "HappyHorse 1.1 Image-to-Video", + "attachment": true, + "reasoning": false, + "tool_call": false, + "temperature": true, + "release_date": "2026-07-17", + "last_updated": "2026-07-17", + "modalities": { + "input": [ + "image", + "text" + ], + "output": [ + "video" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 0, + "output": 0 + } + }, + { + "id": "alibaba-token-plan/happyhorse-1.1-r2v", + "name": "HappyHorse 1.1 Reference-to-Video", + "attachment": true, + "reasoning": false, + "tool_call": false, + "temperature": true, + "release_date": "2026-07-17", + "last_updated": "2026-07-17", + "modalities": { + "input": [ + "image", + "text" + ], + "output": [ + "video" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 0, + "output": 0 + } + }, + { + "id": "alibaba-token-plan/happyhorse-1.1-t2v", + "name": "HappyHorse 1.1 Text-to-Video", + "attachment": false, + "reasoning": false, + "tool_call": false, + "temperature": true, + "release_date": "2026-07-17", + "last_updated": "2026-07-17", + "modalities": { + "input": [ + "text" + ], + "output": [ + "video" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 0, + "output": 0 + } + }, { "id": "alibaba-token-plan/kimi-k2.5", "name": "Kimi K2.5", @@ -11930,7 +12397,7 @@ }, "limit": { "context": 262144, - "output": 32768 + "output": 98304 } }, { @@ -11963,7 +12430,7 @@ }, "limit": { "context": 262144, - "output": 16384 + "output": 262144 } }, { @@ -12091,7 +12558,7 @@ "id": "alibaba-token-plan/qwen3.6-plus", "name": "Qwen3.6 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -12147,14 +12614,14 @@ }, "limit": { "context": 1000000, - "output": 65536 + "output": 131072 } }, { "id": "alibaba-token-plan/qwen3.7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -12164,7 +12631,8 @@ "modalities": { "input": [ "text", - "image" + "image", + "video" ], "output": [ "text" @@ -12179,7 +12647,7 @@ }, "limit": { "context": 1000000, - "output": 64000 + "output": 65536 } }, { @@ -17457,9 +17925,9 @@ }, "open_weights": true, "cost": { - "input": 0.5, - "output": 0.8, - "cache_read": 0.2, + "input": 0.14, + "output": 0.28, + "cache_read": 0.028, "cache_write": 0.0 }, "limit": { @@ -17522,9 +17990,9 @@ }, "open_weights": true, "cost": { - "input": 0.85, - "output": 3.8, - "cache_read": 0.17, + "input": 0.73, + "output": 3.5, + "cache_read": 0.15, "cache_write": 0.0 }, "limit": { @@ -19351,7 +19819,7 @@ "id": "auriko/kimi-k2.5", "name": "Kimi K2.5", "family": "kimi-k2", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": false, @@ -19472,7 +19940,7 @@ "id": "auriko/qwen-3.6-plus", "name": "Qwen3.6 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -19819,64 +20287,6 @@ "output": 8192 } }, - { - "id": "azure-cognitive-services/cohere-command-r-08", - "name": "Command R", - "family": "command-r", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-06-01", - "release_date": "2024-08-30", - "last_updated": "2024-08-30", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.15, - "output": 0.6 - }, - "limit": { - "context": 128000, - "output": 4000 - } - }, - { - "id": "azure-cognitive-services/cohere-command-r-plus-08", - "name": "Command R+", - "family": "command-r", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-06-01", - "release_date": "2024-08-30", - "last_updated": "2024-08-30", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 2.5, - "output": 10.0 - }, - "limit": { - "context": 128000, - "output": 4000 - } - }, { "id": "azure-cognitive-services/cohere-embed-v-4.0", "name": "Embed v4", @@ -19991,64 +20401,6 @@ "output": 163840 } }, - { - "id": "azure-cognitive-services/deepseek-v3", - "name": "DeepSeek-V3-0324", - "family": "deepseek", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-07", - "release_date": "2025-03-24", - "last_updated": "2025-03-24", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 1.14, - "output": 4.56 - }, - "limit": { - "context": 131072, - "output": 131072 - } - }, - { - "id": "azure-cognitive-services/deepseek-v3.1", - "name": "DeepSeek-V3.1", - "family": "deepseek", - "attachment": false, - "reasoning": true, - "tool_call": true, - "temperature": true, - "knowledge": "2024-07", - "release_date": "2025-08-21", - "last_updated": "2025-08-21", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.56, - "output": 1.68 - }, - "limit": { - "context": 131072, - "output": 131072 - } - }, { "id": "azure-cognitive-services/deepseek-v3.2", "name": "DeepSeek-V3.2", @@ -20165,64 +20517,6 @@ "output": 4096 } }, - { - "id": "azure-cognitive-services/gpt-4", - "name": "GPT-4", - "family": "gpt", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-11", - "release_date": "2023-03-14", - "last_updated": "2023-03-14", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 60.0, - "output": 120.0 - }, - "limit": { - "context": 8192, - "output": 8192 - } - }, - { - "id": "azure-cognitive-services/gpt-4-32k", - "name": "GPT-4 32K", - "family": "gpt", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-11", - "release_date": "2023-03-14", - "last_updated": "2023-03-14", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 60.0, - "output": 120.0 - }, - "limit": { - "context": 32768, - "output": 32768 - } - }, { "id": "azure-cognitive-services/gpt-4-turbo", "name": "GPT-4 Turbo", @@ -20469,37 +20763,6 @@ "output": 128000 } }, - { - "id": "azure-cognitive-services/gpt-5-chat", - "name": "GPT-5 Chat", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": false, - "temperature": false, - "knowledge": "2024-10-24", - "release_date": "2025-08-07", - "last_updated": "2025-08-07", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.25, - "output": 10.0, - "cache_read": 0.13 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, { "id": "azure-cognitive-services/gpt-5-codex", "name": "GPT-5-Codex", @@ -20657,40 +20920,6 @@ "output": 128000 } }, - { - "id": "azure-cognitive-services/gpt-5.1-chat", - "name": "GPT-5.1 Chat", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2024-09-30", - "release_date": "2025-11-14", - "last_updated": "2025-11-14", - "modalities": { - "input": [ - "text", - "image", - "audio" - ], - "output": [ - "text", - "image", - "audio" - ] - }, - "open_weights": false, - "cost": { - "input": 1.25, - "output": 10.0, - "cache_read": 0.125 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, { "id": "azure-cognitive-services/gpt-5.1-codex", "name": "GPT-5.1 Codex", @@ -20787,37 +21016,6 @@ "output": 128000 } }, - { - "id": "azure-cognitive-services/gpt-5.2-chat", - "name": "GPT-5.2 Chat", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2025-08-31", - "release_date": "2025-12-11", - "last_updated": "2025-12-11", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.75, - "output": 14.0, - "cache_read": 0.175 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, { "id": "azure-cognitive-services/gpt-5.2-codex", "name": "GPT-5.2 Codex", @@ -21165,72 +21363,11 @@ "output": 16384 } }, - { - "id": "azure-cognitive-services/grok-4-fast-reasoning", - "name": "Grok 4 Fast (Reasoning)", - "family": "grok", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": true, - "knowledge": "2025-07", - "release_date": "2025-09-19", - "last_updated": "2025-09-19", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 0.2, - "output": 0.5, - "cache_read": 0.05 - }, - "limit": { - "context": 2000000, - "output": 30000 - } - }, - { - "id": "azure-cognitive-services/kimi-k2-thinking", - "name": "Kimi K2 Thinking", - "family": "kimi-thinking", - "attachment": false, - "reasoning": true, - "tool_call": true, - "temperature": true, - "knowledge": "2024-08", - "release_date": "2025-11-06", - "last_updated": "2025-12-02", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.6, - "output": 2.5, - "cache_read": 0.15 - }, - "limit": { - "context": 262144, - "output": 262144 - } - }, { "id": "azure-cognitive-services/kimi-k2.5", "name": "Kimi K2.5", "family": "kimi-k2", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -21286,66 +21423,6 @@ "output": 262144 } }, - { - "id": "azure-cognitive-services/llama-3.2-11b-vision-instruct", - "name": "Llama-3.2-11B-Vision-Instruct", - "family": "llama", - "attachment": true, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-09-25", - "last_updated": "2024-09-25", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.37, - "output": 0.37 - }, - "limit": { - "context": 128000, - "output": 8192 - } - }, - { - "id": "azure-cognitive-services/llama-3.2-90b-vision-instruct", - "name": "Llama-3.2-90B-Vision-Instruct", - "family": "llama", - "attachment": true, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-09-25", - "last_updated": "2024-09-25", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 2.04, - "output": 2.04 - }, - "limit": { - "context": 128000, - "output": 8192 - } - }, { "id": "azure-cognitive-services/llama-3.3-70b-instruct", "name": "Llama-3.3-70B-Instruct", @@ -21435,151 +21512,6 @@ "output": 8192 } }, - { - "id": "azure-cognitive-services/meta-llama-3-70b-instruct", - "name": "Meta-Llama-3-70B-Instruct", - "family": "llama", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-04-18", - "last_updated": "2024-04-18", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 2.68, - "output": 3.54 - }, - "limit": { - "context": 8192, - "output": 2048 - } - }, - { - "id": "azure-cognitive-services/meta-llama-3-8b-instruct", - "name": "Meta-Llama-3-8B-Instruct", - "family": "llama", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-04-18", - "last_updated": "2024-04-18", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.3, - "output": 0.61 - }, - "limit": { - "context": 8192, - "output": 2048 - } - }, - { - "id": "azure-cognitive-services/meta-llama-3.1-405b-instruct", - "name": "Meta-Llama-3.1-405B-Instruct", - "family": "llama", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-07-23", - "last_updated": "2024-07-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 5.33, - "output": 16.0 - }, - "limit": { - "context": 128000, - "output": 32768 - } - }, - { - "id": "azure-cognitive-services/meta-llama-3.1-70b-instruct", - "name": "Meta-Llama-3.1-70B-Instruct", - "family": "llama", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-07-23", - "last_updated": "2024-07-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 2.68, - "output": 3.54 - }, - "limit": { - "context": 128000, - "output": 32768 - } - }, - { - "id": "azure-cognitive-services/meta-llama-3.1-8b-instruct", - "name": "Meta-Llama-3.1-8B-Instruct", - "family": "llama", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-07-23", - "last_updated": "2024-07-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.3, - "output": 0.61 - }, - "limit": { - "context": 128000, - "output": 32768 - } - }, { "id": "azure-cognitive-services/ministral-3b", "name": "Ministral 3B", @@ -21609,35 +21541,6 @@ "output": 8192 } }, - { - "id": "azure-cognitive-services/mistral-large", - "name": "Mistral Large 24.11", - "family": "mistral-large", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-09", - "release_date": "2024-11-01", - "last_updated": "2024-11-01", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 2.0, - "output": 6.0 - }, - "limit": { - "context": 128000, - "output": 32768 - } - }, { "id": "azure-cognitive-services/mistral-medium", "name": "Mistral Medium 3", @@ -21668,35 +21571,6 @@ "output": 128000 } }, - { - "id": "azure-cognitive-services/mistral-nemo", - "name": "Mistral Nemo", - "family": "mistral-nemo", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-07", - "release_date": "2024-07-18", - "last_updated": "2024-07-18", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.15, - "output": 0.15 - }, - "limit": { - "context": 128000, - "output": 128000 - } - }, { "id": "azure-cognitive-services/mistral-small", "name": "Mistral Small 3.1", @@ -21786,36 +21660,6 @@ "output": 100000 } }, - { - "id": "azure-cognitive-services/o1-mini", - "name": "o1-mini", - "family": "o-mini", - "attachment": false, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2023-09", - "release_date": "2024-09-12", - "last_updated": "2024-09-12", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.1, - "output": 4.4, - "cache_read": 0.55 - }, - "limit": { - "context": 128000, - "output": 65536 - } - }, { "id": "azure-cognitive-services/o3", "name": "o3", @@ -21908,238 +21752,6 @@ "output": 100000 } }, - { - "id": "azure-cognitive-services/phi-3-medium-128k-instruct", - "name": "Phi-3-medium-instruct (128k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.17, - "output": 0.68 - }, - "limit": { - "context": 128000, - "output": 4096 - } - }, - { - "id": "azure-cognitive-services/phi-3-medium-4k-instruct", - "name": "Phi-3-medium-instruct (4k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.17, - "output": 0.68 - }, - "limit": { - "context": 4096, - "output": 1024 - } - }, - { - "id": "azure-cognitive-services/phi-3-mini-128k-instruct", - "name": "Phi-3-mini-instruct (128k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.13, - "output": 0.52 - }, - "limit": { - "context": 128000, - "output": 4096 - } - }, - { - "id": "azure-cognitive-services/phi-3-mini-4k-instruct", - "name": "Phi-3-mini-instruct (4k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.13, - "output": 0.52 - }, - "limit": { - "context": 4096, - "output": 1024 - } - }, - { - "id": "azure-cognitive-services/phi-3-small-128k-instruct", - "name": "Phi-3-small-instruct (128k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.15, - "output": 0.6 - }, - "limit": { - "context": 128000, - "output": 4096 - } - }, - { - "id": "azure-cognitive-services/phi-3-small-8k-instruct", - "name": "Phi-3-small-instruct (8k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.15, - "output": 0.6 - }, - "limit": { - "context": 8192, - "output": 2048 - } - }, - { - "id": "azure-cognitive-services/phi-3.5-mini-instruct", - "name": "Phi-3.5-mini-instruct", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-08-20", - "last_updated": "2024-08-20", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.13, - "output": 0.52 - }, - "limit": { - "context": 128000, - "output": 4096 - } - }, - { - "id": "azure-cognitive-services/phi-3.5-moe-instruct", - "name": "Phi-3.5-MoE-instruct", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-08-20", - "last_updated": "2024-08-20", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.16, - "output": 0.64 - }, - "limit": { - "context": 128000, - "output": 4096 - } - }, { "id": "azure-cognitive-services/phi-4", "name": "Phi-4", @@ -22815,64 +22427,6 @@ "output": 8192 } }, - { - "id": "azure/cohere-command-r-08", - "name": "Command R", - "family": "command-r", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-06-01", - "release_date": "2024-08-30", - "last_updated": "2024-08-30", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.15, - "output": 0.6 - }, - "limit": { - "context": 128000, - "output": 4000 - } - }, - { - "id": "azure/cohere-command-r-plus-08", - "name": "Command R+", - "family": "command-r", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-06-01", - "release_date": "2024-08-30", - "last_updated": "2024-08-30", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 2.5, - "output": 10.0 - }, - "limit": { - "context": 128000, - "output": 4000 - } - }, { "id": "azure/cohere-embed-v-4.0", "name": "Embed v4", @@ -22987,64 +22541,6 @@ "output": 163840 } }, - { - "id": "azure/deepseek-v3", - "name": "DeepSeek-V3-0324", - "family": "deepseek", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-07", - "release_date": "2025-03-24", - "last_updated": "2025-03-24", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 1.14, - "output": 4.56 - }, - "limit": { - "context": 131072, - "output": 131072 - } - }, - { - "id": "azure/deepseek-v3.1", - "name": "DeepSeek-V3.1", - "family": "deepseek", - "attachment": false, - "reasoning": true, - "tool_call": true, - "temperature": true, - "knowledge": "2024-07", - "release_date": "2025-08-21", - "last_updated": "2025-08-21", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.56, - "output": 1.68 - }, - "limit": { - "context": 131072, - "output": 131072 - } - }, { "id": "azure/deepseek-v3.2", "name": "DeepSeek-V3.2", @@ -23219,64 +22715,6 @@ "output": 4096 } }, - { - "id": "azure/gpt-4", - "name": "GPT-4", - "family": "gpt", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-11", - "release_date": "2023-03-14", - "last_updated": "2023-03-14", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 60.0, - "output": 120.0 - }, - "limit": { - "context": 8192, - "output": 8192 - } - }, - { - "id": "azure/gpt-4-32k", - "name": "GPT-4 32K", - "family": "gpt", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-11", - "release_date": "2023-03-14", - "last_updated": "2023-03-14", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 60.0, - "output": 120.0 - }, - "limit": { - "context": 32768, - "output": 32768 - } - }, { "id": "azure/gpt-4-turbo", "name": "GPT-4 Turbo", @@ -23523,37 +22961,6 @@ "output": 128000 } }, - { - "id": "azure/gpt-5-chat", - "name": "GPT-5 Chat", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": false, - "temperature": false, - "knowledge": "2024-10-24", - "release_date": "2025-08-07", - "last_updated": "2025-08-07", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.25, - "output": 10.0, - "cache_read": 0.13 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, { "id": "azure/gpt-5-codex", "name": "GPT-5-Codex", @@ -23711,40 +23118,6 @@ "output": 128000 } }, - { - "id": "azure/gpt-5.1-chat", - "name": "GPT-5.1 Chat", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2024-09-30", - "release_date": "2025-11-14", - "last_updated": "2025-11-14", - "modalities": { - "input": [ - "text", - "image", - "audio" - ], - "output": [ - "text", - "image", - "audio" - ] - }, - "open_weights": false, - "cost": { - "input": 1.25, - "output": 10.0, - "cache_read": 0.125 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, { "id": "azure/gpt-5.1-codex", "name": "GPT-5.1 Codex", @@ -23872,37 +23245,6 @@ "output": 128000 } }, - { - "id": "azure/gpt-5.2-chat", - "name": "GPT-5.2 Chat", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2025-08-31", - "release_date": "2025-12-11", - "last_updated": "2025-12-11", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.75, - "output": 14.0, - "cache_read": 0.175 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, { "id": "azure/gpt-5.2-codex", "name": "GPT-5.2 Codex", @@ -23934,37 +23276,6 @@ "output": 128000 } }, - { - "id": "azure/gpt-5.3-chat", - "name": "GPT-5.3 Chat", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2025-08-31", - "release_date": "2026-03-03", - "last_updated": "2026-03-03", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.75, - "output": 14.0, - "cache_read": 0.175 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, { "id": "azure/gpt-5.3-codex", "name": "GPT-5.3 Codex", @@ -24430,37 +23741,6 @@ "output": 8192 } }, - { - "id": "azure/grok-4-fast-reasoning", - "name": "Grok 4 Fast (Reasoning)", - "family": "grok", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": true, - "knowledge": "2025-07", - "release_date": "2025-09-19", - "last_updated": "2025-09-19", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 0.2, - "output": 0.5, - "cache_read": 0.05 - }, - "limit": { - "context": 2000000, - "output": 30000 - } - }, { "id": "azure/grok-4.1-fast-non-reasoning", "name": "Grok 4.1 Fast (Non-Reasoning)", @@ -24521,36 +23801,6 @@ "output": 8192 } }, - { - "id": "azure/kimi-k2-thinking", - "name": "Kimi K2 Thinking", - "family": "kimi-thinking", - "attachment": false, - "reasoning": true, - "tool_call": true, - "temperature": true, - "knowledge": "2024-08", - "release_date": "2025-11-06", - "last_updated": "2025-12-02", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.6, - "output": 2.5, - "cache_read": 0.15 - }, - "limit": { - "context": 262144, - "output": 262144 - } - }, { "id": "azure/kimi-k2.5", "name": "Kimi K2.5", @@ -24611,66 +23861,6 @@ "output": 262144 } }, - { - "id": "azure/llama-3.2-11b-vision-instruct", - "name": "Llama-3.2-11B-Vision-Instruct", - "family": "llama", - "attachment": true, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-09-25", - "last_updated": "2024-09-25", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.37, - "output": 0.37 - }, - "limit": { - "context": 128000, - "output": 8192 - } - }, - { - "id": "azure/llama-3.2-90b-vision-instruct", - "name": "Llama-3.2-90B-Vision-Instruct", - "family": "llama", - "attachment": true, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-09-25", - "last_updated": "2024-09-25", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 2.04, - "output": 2.04 - }, - "limit": { - "context": 128000, - "output": 8192 - } - }, { "id": "azure/llama-3.3-70b-instruct", "name": "Llama-3.3-70B-Instruct", @@ -24760,151 +23950,6 @@ "output": 8192 } }, - { - "id": "azure/meta-llama-3-70b-instruct", - "name": "Meta-Llama-3-70B-Instruct", - "family": "llama", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-04-18", - "last_updated": "2024-04-18", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 2.68, - "output": 3.54 - }, - "limit": { - "context": 8192, - "output": 2048 - } - }, - { - "id": "azure/meta-llama-3-8b-instruct", - "name": "Meta-Llama-3-8B-Instruct", - "family": "llama", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-04-18", - "last_updated": "2024-04-18", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.3, - "output": 0.61 - }, - "limit": { - "context": 8192, - "output": 2048 - } - }, - { - "id": "azure/meta-llama-3.1-405b-instruct", - "name": "Meta-Llama-3.1-405B-Instruct", - "family": "llama", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-07-23", - "last_updated": "2024-07-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 5.33, - "output": 16.0 - }, - "limit": { - "context": 128000, - "output": 32768 - } - }, - { - "id": "azure/meta-llama-3.1-70b-instruct", - "name": "Meta-Llama-3.1-70B-Instruct", - "family": "llama", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-07-23", - "last_updated": "2024-07-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 2.68, - "output": 3.54 - }, - "limit": { - "context": 128000, - "output": 32768 - } - }, - { - "id": "azure/meta-llama-3.1-8b-instruct", - "name": "Meta-Llama-3.1-8B-Instruct", - "family": "llama", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2023-12", - "release_date": "2024-07-23", - "last_updated": "2024-07-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.3, - "output": 0.61 - }, - "limit": { - "context": 128000, - "output": 32768 - } - }, { "id": "azure/ministral-3b", "name": "Ministral 3B", @@ -24934,35 +23979,6 @@ "output": 8192 } }, - { - "id": "azure/mistral-large", - "name": "Mistral Large 24.11", - "family": "mistral-large", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-09", - "release_date": "2024-11-01", - "last_updated": "2024-11-01", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 2.0, - "output": 6.0 - }, - "limit": { - "context": 128000, - "output": 32768 - } - }, { "id": "azure/mistral-medium", "name": "Mistral Medium 3", @@ -24993,35 +24009,6 @@ "output": 128000 } }, - { - "id": "azure/mistral-nemo", - "name": "Mistral Nemo", - "family": "mistral-nemo", - "attachment": false, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-07", - "release_date": "2024-07-18", - "last_updated": "2024-07-18", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.15, - "output": 0.15 - }, - "limit": { - "context": 128000, - "output": 128000 - } - }, { "id": "azure/mistral-small", "name": "Mistral Small 3.1", @@ -25111,36 +24098,6 @@ "output": 100000 } }, - { - "id": "azure/o1-mini", - "name": "o1-mini", - "family": "o-mini", - "attachment": false, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2023-09", - "release_date": "2024-09-12", - "last_updated": "2024-09-12", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.1, - "output": 4.4, - "cache_read": 0.55 - }, - "limit": { - "context": 128000, - "output": 65536 - } - }, { "id": "azure/o3", "name": "o3", @@ -25233,238 +24190,6 @@ "output": 100000 } }, - { - "id": "azure/phi-3-medium-128k-instruct", - "name": "Phi-3-medium-instruct (128k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.17, - "output": 0.68 - }, - "limit": { - "context": 128000, - "output": 4096 - } - }, - { - "id": "azure/phi-3-medium-4k-instruct", - "name": "Phi-3-medium-instruct (4k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.17, - "output": 0.68 - }, - "limit": { - "context": 4096, - "output": 1024 - } - }, - { - "id": "azure/phi-3-mini-128k-instruct", - "name": "Phi-3-mini-instruct (128k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.13, - "output": 0.52 - }, - "limit": { - "context": 128000, - "output": 4096 - } - }, - { - "id": "azure/phi-3-mini-4k-instruct", - "name": "Phi-3-mini-instruct (4k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.13, - "output": 0.52 - }, - "limit": { - "context": 4096, - "output": 1024 - } - }, - { - "id": "azure/phi-3-small-128k-instruct", - "name": "Phi-3-small-instruct (128k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.15, - "output": 0.6 - }, - "limit": { - "context": 128000, - "output": 4096 - } - }, - { - "id": "azure/phi-3-small-8k-instruct", - "name": "Phi-3-small-instruct (8k)", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-04-23", - "last_updated": "2024-04-23", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.15, - "output": 0.6 - }, - "limit": { - "context": 8192, - "output": 2048 - } - }, - { - "id": "azure/phi-3.5-mini-instruct", - "name": "Phi-3.5-mini-instruct", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-08-20", - "last_updated": "2024-08-20", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.13, - "output": 0.52 - }, - "limit": { - "context": 128000, - "output": 4096 - } - }, - { - "id": "azure/phi-3.5-moe-instruct", - "name": "Phi-3.5-MoE-instruct", - "family": "phi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2023-10", - "release_date": "2024-08-20", - "last_updated": "2024-08-20", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.16, - "output": 0.64 - }, - "limit": { - "context": 128000, - "output": 4096 - } - }, { "id": "azure/phi-4", "name": "Phi-4", @@ -25960,6 +24685,35 @@ "output": 262000 } }, + { + "id": "baseten/moonshotai/Kimi-K3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-16", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 3.0, + "output": 15.0 + }, + "limit": { + "context": 1048576, + "output": 262144 + } + }, { "id": "baseten/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B", "name": "Nemotron Ultra", @@ -26191,8 +24945,37 @@ "cache_read": 0.3 }, "limit": { - "context": 256000, - "output": 256000 + "context": 524288, + "output": 262144 + } + }, + { + "id": "baseten/zai-org/GLM-5.2-Fast", + "name": "GLM 5.2 Fast", + "family": "glm", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-06-13", + "last_updated": "2026-06-13", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 2.1, + "output": 6.6, + "cache_read": 0.21 + }, + "limit": { + "context": 524288, + "output": 262144 } }, { @@ -26347,6 +25130,35 @@ "output": 262144 } }, + { + "id": "berget/moonshotai/Kimi-K3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-27", + "last_updated": "2026-07-28", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 3.0, + "output": 15.0 + }, + "limit": { + "context": 327680, + "output": 32768 + } + }, { "id": "berget/openai/gpt-oss-120b", "name": "GPT-OSS-120B", @@ -26598,6 +25410,35 @@ "output": 65536 } }, + { + "id": "chutes/Nemotron-3-Nano-Omni-30B-TEE", + "name": "Nemotron 3 Nano Omni 30B TEE", + "family": "nemotron", + "attachment": false, + "reasoning": false, + "tool_call": false, + "temperature": true, + "release_date": "2026-07-23", + "last_updated": "2026-07-23", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0245, + "output": 0.0978, + "cache_read": 0.01225 + }, + "limit": { + "context": 131072, + "output": 0 + } + }, { "id": "chutes/Qwen/Qwen3-235B-A22B-Thinking-2507-TEE", "name": "Qwen3 235B A22B Thinking 2507 TEE", @@ -26949,9 +25790,9 @@ }, "open_weights": true, "cost": { - "input": 1.4, - "output": 4.4, - "cache_read": 0.7 + "input": 1.25, + "output": 3.95, + "cache_read": 0.625 }, "limit": { "context": 1048576, @@ -27519,6 +26360,37 @@ "output": 262144 } }, + { + "id": "cline-pass/cline-pass/kimi-k3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 3.0, + "output": 15.0, + "cache_read": 0.3 + }, + "limit": { + "context": 1048576, + "output": 131072 + } + }, { "id": "cline-pass/cline-pass/mimo-v2.5", "name": "MiMo-V2.5", @@ -27647,7 +26519,7 @@ "id": "cline-pass/cline-pass/qwen3.7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -27657,7 +26529,8 @@ "modalities": { "input": [ "text", - "image" + "image", + "video" ], "output": [ "text" @@ -28246,6 +27119,39 @@ "output": 128000 } }, + { + "id": "cloudflare-ai-gateway/anthropic/claude-opus-5", + "name": "Claude Opus 5", + "family": "claude-opus", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-05", + "release_date": "2026-07-24", + "last_updated": "2026-07-24", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 5.0, + "output": 25.0, + "cache_read": 0.5, + "cache_write": 6.25 + }, + "limit": { + "context": 1000000, + "output": 128000 + } + }, { "id": "cloudflare-ai-gateway/anthropic/claude-sonnet-4", "name": "Claude Sonnet 4 (latest)", @@ -32159,6 +31065,35 @@ "output": 128000 } }, + { + "id": "cortecs/hy3", + "name": "Hy3", + "family": "Hy", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-06", + "last_updated": "2026-07-06", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.41, + "output": 1.025, + "cache_read": 0.103 + }, + "limit": { + "context": 262144, + "output": 64000 + } + }, { "id": "cortecs/intellect-3", "name": "INTELLECT 3", @@ -33365,7 +32300,7 @@ "id": "crof/kimi-k2.5", "name": "Kimi K2.5", "family": "kimi-k2", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": false, @@ -33488,6 +32423,37 @@ "output": 262144 } }, + { + "id": "crof/kimi-k3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 2.0, + "output": 8.0, + "cache_read": 0.25 + }, + "limit": { + "context": 1000000, + "output": 262144 + } + }, { "id": "crof/mimo-v2.5-pro", "name": "MiMo-V2.5-Pro", @@ -33772,6 +32738,38 @@ "output": 128000 } }, + { + "id": "crossmodel/anthropic/claude-opus-5", + "name": "Claude Opus 5", + "family": "claude-opus", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-05", + "release_date": "2026-07-24", + "last_updated": "2026-07-24", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 5.0, + "output": 25.0, + "cache_read": 0.5, + "cache_write": 6.25 + }, + "limit": { + "context": 1000000, + "output": 128000 + } + }, { "id": "crossmodel/anthropic/claude-sonnet-4.6", "name": "Claude Sonnet 4.6", @@ -34168,7 +33166,7 @@ "id": "crossmodel/moonshot/kimi-k2.5", "name": "Kimi K2.5", "family": "kimi-k2", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": false, @@ -34614,7 +33612,7 @@ "id": "crossmodel/qwen/qwen3.6-plus", "name": "Qwen3.6 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -34677,7 +33675,7 @@ "id": "crossmodel/qwen/qwen3.7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -34706,6 +33704,36 @@ "output": 64000 } }, + { + "id": "crossmodel/tencent/hy3", + "name": "Hy3", + "family": "Hy", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-06", + "last_updated": "2026-07-06", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.16, + "output": 0.64, + "cache_read": 0.04, + "cache_write": 0.16 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, { "id": "crossmodel/tencent/hy3-preview", "name": "Hy3 preview", @@ -35305,7 +34333,7 @@ "id": "daoxe/kimi-k2.5", "name": "Kimi K2.5", "family": "kimi-k2", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": false, @@ -37169,6 +36197,36 @@ "output": 262144 } }, + { + "id": "deepinfra/moonshotai/Kimi-K3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 2.7, + "output": 13.5, + "cache_read": 0.27 + }, + "limit": { + "context": 1048576, + "output": 131072 + } + }, { "id": "deepinfra/nvidia/Llama-3.3-Nemotron-Super-49B-v1.5", "name": "Llama 3.3 Nemotron Super 49B v1.5", @@ -40105,6 +39163,34 @@ "output": 128000 } }, + { + "id": "dinference/glm-5.2", + "name": "GLM-5.2", + "family": "glm", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-06-13", + "last_updated": "2026-06-13", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 1.25, + "output": 3.89 + }, + "limit": { + "context": 1000000, + "output": 128000 + } + }, { "id": "dinference/gpt-oss-120b", "name": "GPT OSS 120B", @@ -40419,36 +39505,6 @@ "output": 393216 } }, - { - "id": "empiriolabs/fugu-ultra", - "name": "Fugu Ultra", - "family": "fugu", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "release_date": "2026-06-15", - "last_updated": "2026-06-15", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 7.5, - "output": 45.0, - "cache_read": 1.5 - }, - "limit": { - "context": 1000000, - "output": 131072 - } - }, { "id": "empiriolabs/gemma-4-26b-a4b", "name": "Gemma 4 26B-A4B", @@ -43332,35 +42388,6 @@ "output": 384000 } }, - { - "id": "fireworks-ai/accounts/fireworks/models/glm-5p1", - "name": "GLM 5.1", - "family": "glm", - "attachment": false, - "reasoning": true, - "tool_call": true, - "temperature": true, - "release_date": "2026-04-01", - "last_updated": "2026-04-01", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 1.4, - "output": 4.4, - "cache_read": 0.26 - }, - "limit": { - "context": 202800, - "output": 131072 - } - }, { "id": "fireworks-ai/accounts/fireworks/models/glm-5p2", "name": "GLM 5.2", @@ -43508,6 +42535,36 @@ "output": 262000 } }, + { + "id": "fireworks-ai/accounts/fireworks/models/kimi-k3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-27", + "last_updated": "2026-07-27", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 3.0, + "output": 15.0, + "cache_read": 0.3 + }, + "limit": { + "context": 1048576, + "output": 131072 + } + }, { "id": "fireworks-ai/accounts/fireworks/models/minimax-m2p7", "name": "MiniMax-M2.7", @@ -43549,7 +42606,9 @@ "last_updated": "2026-06-12", "modalities": { "input": [ - "text" + "text", + "image", + "video" ], "output": [ "text" @@ -43596,35 +42655,6 @@ "output": 65536 } }, - { - "id": "fireworks-ai/accounts/fireworks/routers/glm-5p1-fast", - "name": "GLM 5.1 Fast", - "family": "glm", - "attachment": false, - "reasoning": true, - "tool_call": true, - "temperature": true, - "release_date": "2026-04-01", - "last_updated": "2026-04-01", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 2.8, - "output": 8.8, - "cache_read": 0.52 - }, - "limit": { - "context": 202800, - "output": 131072 - } - }, { "id": "fireworks-ai/accounts/fireworks/routers/glm-5p2-fast", "name": "GLM 5.2 Fast", @@ -43744,6 +42774,36 @@ "output": 262000 } }, + { + "id": "fireworks-ai/accounts/fireworks/routers/kimi-k3-fast", + "name": "Kimi K3 Fast", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-27", + "last_updated": "2026-07-27", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 4.5, + "output": 22.5, + "cache_read": 0.45 + }, + "limit": { + "context": 1048576, + "output": 131072 + } + }, { "id": "freemodel/claude-fable-5", "name": "Claude Fable 5", @@ -48190,6 +47250,39 @@ "output": 128000 } }, + { + "id": "gitlab/duo-chat-opus-5", + "name": "Agentic Chat (Claude Opus 5)", + "family": "claude-opus", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-05", + "release_date": "2026-07-24", + "last_updated": "2026-07-24", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0, + "cache_read": 0.0, + "cache_write": 0.0 + }, + "limit": { + "context": 1000000, + "output": 128000 + } + }, { "id": "gitlab/duo-chat-sonnet-4.5", "name": "Agentic Chat (Claude Sonnet 4.5)", @@ -48680,39 +47773,6 @@ "output": 131072 } }, - { - "id": "google-vertex-anthropic/claude-3.5-haiku", - "name": "Claude Haiku 3.5", - "family": "claude-haiku", - "attachment": true, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-07-31", - "release_date": "2024-10-22", - "last_updated": "2024-10-22", - "modalities": { - "input": [ - "text", - "image", - "pdf" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 0.8, - "output": 4.0, - "cache_read": 0.08, - "cache_write": 1.0 - }, - "limit": { - "context": 200000, - "output": 8192 - } - }, { "id": "google-vertex-anthropic/claude-haiku-4.5", "name": "Claude Haiku 4.5", @@ -49109,39 +48169,6 @@ "output": 128000 } }, - { - "id": "google-vertex/claude-3.5-haiku", - "name": "Claude Haiku 3.5", - "family": "claude-haiku", - "attachment": true, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-07-31", - "release_date": "2024-10-22", - "last_updated": "2024-10-22", - "modalities": { - "input": [ - "text", - "image", - "pdf" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 0.8, - "output": 4.0, - "cache_read": 0.08, - "cache_write": 1.0 - }, - "limit": { - "context": 200000, - "output": 8192 - } - }, { "id": "google-vertex/claude-haiku-4.5", "name": "Claude Haiku 4.5", @@ -50057,6 +49084,74 @@ "output": 65536 } }, + { + "id": "google-vertex/gemini-3.5-flash-lite", + "name": "Gemini 3.5 Flash Lite", + "family": "gemini-flash-lite", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2026-03", + "release_date": "2026-07-21", + "last_updated": "2026-07-21", + "modalities": { + "input": [ + "text", + "image", + "video", + "audio", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.3, + "output": 2.5, + "cache_read": 0.03 + }, + "limit": { + "context": 1048576, + "output": 65536 + } + }, + { + "id": "google-vertex/gemini-3.6-flash", + "name": "Gemini 3.6 Flash", + "family": "gemini-flash", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2026-03", + "release_date": "2026-07-21", + "last_updated": "2026-07-21", + "modalities": { + "input": [ + "text", + "image", + "video", + "audio", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.5, + "output": 7.5, + "cache_read": 0.15 + }, + "limit": { + "context": 1048576, + "output": 65536 + } + }, { "id": "google-vertex/gemini-embedding-001", "name": "Gemini Embedding 001", @@ -50385,6 +49480,76 @@ "output": 131072 } }, + { + "id": "google/deep-research-max-preview-04", + "name": "Deep Research Max Preview (Apr-21-2026)", + "family": "gemini-pro", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-04-21", + "last_updated": "2026-04-21", + "modalities": { + "input": [ + "text", + "image", + "video", + "audio", + "pdf" + ], + "output": [ + "text", + "image" + ] + }, + "open_weights": false, + "cost": { + "input": 2.0, + "output": 12.0, + "cache_read": 0.2 + }, + "limit": { + "context": 131072, + "output": 65536 + } + }, + { + "id": "google/deep-research-preview-04", + "name": "Deep Research Preview (Apr-21-2026)", + "family": "gemini-pro", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-04-21", + "last_updated": "2026-04-21", + "modalities": { + "input": [ + "text", + "image", + "video", + "audio", + "pdf" + ], + "output": [ + "text", + "image" + ] + }, + "open_weights": false, + "cost": { + "input": 2.0, + "output": 12.0, + "cache_read": 0.2 + }, + "limit": { + "context": 131072, + "output": 65536 + } + }, { "id": "google/gemini-2.0-flash", "name": "Gemini 2.0 Flash", @@ -50452,6 +49617,36 @@ "output": 8192 } }, + { + "id": "google/gemini-2.5-computer-use-preview-10", + "name": "Gemini 2.5 Computer Use Preview 10-2025", + "family": "gemini-pro", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2025-10-07", + "last_updated": "2025-10-07", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.25, + "output": 10.0 + }, + "limit": { + "context": 131072, + "output": 65536 + } + }, { "id": "google/gemini-2.5-flash", "name": "Gemini 2.5 Flash", @@ -50678,6 +49873,37 @@ "output": 65536 } }, + { + "id": "google/gemini-3-pro-image", + "name": "Nano Banana Pro", + "family": "gemini-pro", + "attachment": true, + "reasoning": true, + "tool_call": false, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-05-28", + "last_updated": "2026-05-28", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text", + "image" + ] + }, + "open_weights": false, + "cost": { + "input": 2.0, + "output": 120.0 + }, + "limit": { + "context": 131072, + "output": 32768 + } + }, { "id": "google/gemini-3-pro-image-preview", "name": "Nano Banana Pro", @@ -50743,6 +49969,39 @@ "output": 65536 } }, + { + "id": "google/gemini-3.1-flash-image", + "name": "Nano Banana 2", + "family": "gemini-flash", + "attachment": true, + "reasoning": true, + "tool_call": false, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-05-28", + "last_updated": "2026-05-28", + "modalities": { + "input": [ + "text", + "image", + "video", + "pdf" + ], + "output": [ + "text", + "image" + ] + }, + "open_weights": false, + "cost": { + "input": 0.5, + "output": 60.0 + }, + "limit": { + "context": 65536, + "output": 65536 + } + }, { "id": "google/gemini-3.1-flash-image-preview", "name": "Nano Banana 2", @@ -50809,6 +50068,37 @@ "output": 65536 } }, + { + "id": "google/gemini-3.1-flash-lite-image", + "name": "Nano Banana 2 Lite", + "family": "gemini-flash-lite", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-06-30", + "last_updated": "2026-06-30", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text", + "image" + ] + }, + "open_weights": false, + "cost": { + "input": 0.25, + "output": 30.0 + }, + "limit": { + "context": 65536, + "output": 65536 + } + }, { "id": "google/gemini-3.1-flash-lite-preview", "name": "Gemini 3.1 Flash Lite Preview", @@ -50843,6 +50133,68 @@ "output": 65536 } }, + { + "id": "google/gemini-3.1-flash-live-preview", + "name": "Gemini 3.1 Flash Live Preview", + "family": "gemini-flash", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-03-26", + "last_updated": "2026-03-26", + "modalities": { + "input": [ + "text", + "image", + "video", + "audio" + ], + "output": [ + "text", + "audio" + ] + }, + "open_weights": false, + "cost": { + "input": 0.75, + "output": 4.5 + }, + "limit": { + "context": 131072, + "output": 65536 + } + }, + { + "id": "google/gemini-3.1-flash-tts-preview", + "name": "Gemini 3.1 Flash TTS Preview", + "family": "gemini-flash", + "attachment": false, + "reasoning": true, + "tool_call": false, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-04-15", + "last_updated": "2026-04-15", + "modalities": { + "input": [ + "text" + ], + "output": [ + "audio" + ] + }, + "open_weights": false, + "cost": { + "input": 1.0, + "output": 20.0 + }, + "limit": { + "context": 8192, + "output": 16384 + } + }, { "id": "google/gemini-3.1-pro-preview", "name": "Gemini 3.1 Pro Preview", @@ -50979,6 +50331,36 @@ "output": 65536 } }, + { + "id": "google/gemini-3.5-live-translate-preview", + "name": "Gemini 3.5 Live Translate Preview", + "family": "gemini-pro", + "attachment": false, + "reasoning": false, + "tool_call": false, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-06-09", + "last_updated": "2026-06-09", + "modalities": { + "input": [ + "audio" + ], + "output": [ + "audio", + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 3.5, + "output": 21.0 + }, + "limit": { + "context": 16384, + "output": 32768 + } + }, { "id": "google/gemini-3.6-flash", "name": "Gemini 3.6 Flash", @@ -51042,6 +50424,39 @@ "output": 1 } }, + { + "id": "google/gemini-embedding-2", + "name": "Gemini Embedding 2", + "family": "gemini", + "attachment": true, + "reasoning": false, + "tool_call": false, + "temperature": false, + "knowledge": "2025-11", + "release_date": "2026-04-22", + "last_updated": "2026-04-22", + "modalities": { + "input": [ + "text", + "image", + "audio", + "video", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.2, + "output": 0.0 + }, + "limit": { + "context": 8192, + "output": 1 + } + }, { "id": "google/gemini-flash", "name": "Gemini Flash Latest", @@ -51140,6 +50555,38 @@ "output": 65536 } }, + { + "id": "google/gemini-robotics-er-1.6-preview", + "name": "Gemini Robotics-ER 1.6 Preview", + "family": "gemini", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-04-14", + "last_updated": "2026-04-14", + "modalities": { + "input": [ + "text", + "image", + "video", + "audio" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.0, + "output": 5.0 + }, + "limit": { + "context": 131072, + "output": 65536 + } + }, { "id": "google/gemma-4-26b-a4b-it", "name": "Gemma 4 26B A4B IT", @@ -51192,6 +50639,145 @@ "output": 32768 } }, + { + "id": "google/lyria-3-clip-preview", + "name": "Lyria 3 Clip Preview", + "family": "lyria", + "attachment": true, + "reasoning": false, + "tool_call": false, + "temperature": true, + "release_date": "2026-03-25", + "last_updated": "2026-03-25", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text", + "audio" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1048576, + "output": 65536 + } + }, + { + "id": "google/lyria-3-pro-preview", + "name": "Lyria 3 Pro Preview", + "family": "lyria", + "attachment": true, + "reasoning": false, + "tool_call": false, + "temperature": true, + "release_date": "2026-03-25", + "last_updated": "2026-03-25", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text", + "audio" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1048576, + "output": 65536 + } + }, + { + "id": "google/veo-3.1-fast-generate-preview", + "name": "Veo 3.1 fast", + "family": "veo", + "attachment": true, + "reasoning": false, + "tool_call": false, + "temperature": false, + "release_date": "2025-10-15", + "last_updated": "2026-01-01", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "video" + ] + }, + "open_weights": false, + "cost": {}, + "limit": { + "context": 480, + "output": 8192 + } + }, + { + "id": "google/veo-3.1-generate-preview", + "name": "Veo 3.1", + "family": "veo", + "attachment": true, + "reasoning": false, + "tool_call": false, + "temperature": false, + "release_date": "2025-10-15", + "last_updated": "2026-01", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "video" + ] + }, + "open_weights": false, + "cost": {}, + "limit": { + "context": 480, + "output": 8192 + } + }, + { + "id": "google/veo-3.1-lite-generate-preview", + "name": "Veo 3.1 lite", + "family": "veo", + "attachment": true, + "reasoning": false, + "tool_call": false, + "temperature": false, + "release_date": "2026-03-31", + "last_updated": "2026-03-31", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "video" + ] + }, + "open_weights": false, + "cost": {}, + "limit": { + "context": 480, + "output": 8192 + } + }, { "id": "groq/canopylabs/orpheus-arabic-saudi", "name": "Canopy Labs Orpheus Arabic Saudi", @@ -54247,6 +53833,36 @@ "output": 4096 } }, + { + "id": "hetzner/Qwen/Qwen3.6-35B-A3B-FP8", + "name": "Qwen3.6 35B A3B FP8", + "family": "qwen", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-04-17", + "last_updated": "2026-04-17", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0, + "cache_read": 0.0 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, { "id": "hpc-ai/anthropic/claude-opus-4.7", "name": "Claude Opus 4.7", @@ -55185,6 +54801,35 @@ "output": 4096 } }, + { + "id": "huggingface/XiaomiMiMo/MiMo-V2.5", + "name": "MiMo-V2.5", + "family": "mimo", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2024-12", + "release_date": "2026-04-22", + "last_updated": "2026-04-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.4, + "output": 2.0 + }, + "limit": { + "context": 262144, + "output": 131072 + } + }, { "id": "huggingface/XiaomiMiMo/MiMo-V2.5-Pro", "name": "MiMo-V2.5-Pro", @@ -55571,6 +55216,35 @@ "output": 262144 } }, + { + "id": "huggingface/moonshotai/Kimi-K3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 3.0, + "output": 15.0 + }, + "limit": { + "context": 1000000, + "output": 131072 + } + }, { "id": "huggingface/openai/gpt-oss-120b", "name": "GPT OSS 120B", @@ -55947,6 +55621,606 @@ "output": 131072 } }, + { + "id": "hyper/deepseek-v4-flash", + "name": "DeepSeek V4 Flash", + "family": "deepseek-flash", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-05", + "release_date": "2026-07-06", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.2, + "output": 0.4, + "cache_read": 0.04 + }, + "limit": { + "context": 1000000, + "output": 384000 + } + }, + { + "id": "hyper/deepseek-v4-pro", + "name": "DeepSeek V4 Pro", + "family": "deepseek-thinking", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-05", + "release_date": "2026-07-06", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 2.4, + "output": 4.8, + "cache_read": 0.2 + }, + "limit": { + "context": 1000000, + "output": 384000 + } + }, + { + "id": "hyper/gemma-4-26b-a4b-it", + "name": "Gemma 4 26B A4B IT", + "family": "gemma", + "attachment": false, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2026-04-30", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.13, + "output": 0.424, + "cache_write": 0.065 + }, + "limit": { + "context": 256000, + "output": 25600 + } + }, + { + "id": "hyper/glm-5", + "name": "GLM-5", + "family": "glm", + "attachment": false, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2026-04-13", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.81, + "output": 2.624, + "cache_write": 0.405 + }, + "limit": { + "context": 202752, + "output": 20275 + } + }, + { + "id": "hyper/glm-5.1", + "name": "GLM-5.1", + "family": "glm", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-06-04", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 1.52432, + "output": 4.79072, + "cache_read": 0.283088 + }, + "limit": { + "context": 202800, + "output": 131072 + } + }, + { + "id": "hyper/glm-5.2", + "name": "GLM-5.2", + "family": "glm", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-06-30", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 1.4, + "output": 4.4, + "cache_read": 0.28 + }, + "limit": { + "context": 1000000, + "output": 128000 + } + }, + { + "id": "hyper/gpt-oss-120b", + "name": "GPT OSS 120B", + "family": "gpt-oss", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-04-13", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.176, + "output": 0.652, + "cache_write": 0.088 + }, + "limit": { + "context": 131072, + "output": 13107 + } + }, + { + "id": "hyper/kimi-k2.5", + "name": "Kimi K2.5", + "family": "kimi-k2", + "attachment": false, + "reasoning": false, + "tool_call": true, + "temperature": false, + "knowledge": "2025-01", + "release_date": "2026-04-13", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.528, + "output": 2.79, + "cache_write": 0.264 + }, + "limit": { + "context": 262144, + "output": 26214 + } + }, + { + "id": "hyper/kimi-k2.6", + "name": "Kimi K2.6", + "family": "kimi-k2", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-07-03", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.95, + "output": 4.0, + "cache_write": 0.16 + }, + "limit": { + "context": 262000, + "output": 262000 + } + }, + { + "id": "hyper/kimi-k2.7-code", + "name": "Kimi K2.7 Code", + "family": "kimi-k2", + "attachment": true, + "reasoning": false, + "tool_call": true, + "temperature": false, + "knowledge": "2025-01", + "release_date": "2026-07-03", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.95, + "output": 4.0, + "cache_read": 0.19 + }, + "limit": { + "context": 256000, + "output": 16000 + } + }, + { + "id": "hyper/llama-3.3-70b-instruct", + "name": "Llama-3.3-70B-Instruct", + "family": "llama", + "attachment": false, + "reasoning": false, + "tool_call": true, + "temperature": true, + "knowledge": "2023-12", + "release_date": "2026-04-30", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.6066, + "output": 1.0386, + "cache_write": 0.3033 + }, + "limit": { + "context": 128000, + "output": 12800 + } + }, + { + "id": "hyper/llama-4-maverick-17b-128e-instruct-fp8", + "name": "Llama 4 Maverick 17B Instruct", + "family": "llama", + "attachment": false, + "reasoning": false, + "tool_call": true, + "temperature": true, + "knowledge": "2024-08", + "release_date": "2026-04-30", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.284, + "output": 0.934, + "cache_write": 0.142 + }, + "limit": { + "context": 430000, + "output": 43000 + } + }, + { + "id": "hyper/minimax-m2.7", + "name": "MiniMax-M2.7", + "family": "minimax", + "attachment": false, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2026-06-05", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.348, + "output": 1.332, + "cache_write": 0.174 + }, + "limit": { + "context": 204800, + "output": 20480 + } + }, + { + "id": "hyper/qwen3-coder-480b-a35b-instruct-int4-mixed-ar", + "name": "Qwen3-Coder 480B-A35B Instruct", + "family": "qwen", + "attachment": false, + "reasoning": false, + "tool_call": true, + "temperature": true, + "knowledge": "2025-04", + "release_date": "2026-04-30", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.569, + "output": 2.135, + "cache_write": 0.2845 + }, + "limit": { + "context": 106000, + "output": 10600 + } + }, + { + "id": "hyper/qwen3-next-80b-a3b-instruct", + "name": "Qwen3-Next 80B-A3B Instruct", + "family": "qwen", + "attachment": false, + "reasoning": false, + "tool_call": true, + "temperature": true, + "knowledge": "2025-04", + "release_date": "2026-04-30", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.1175, + "output": 1.136, + "cache_write": 0.05875 + }, + "limit": { + "context": 262144, + "output": 26214 + } + }, + { + "id": "hyper/qwen3.6-flash", + "name": "Qwen3.6 Flash", + "family": "qwen3.6", + "attachment": true, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2026-05-20", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.0, + "output": 4.0, + "cache_read": 0.1, + "cache_write": 1.25 + }, + "limit": { + "context": 1000000, + "output": 64000 + } + }, + { + "id": "hyper/qwen3.6-max", + "name": "Qwen3.6 Max Preview", + "family": "qwen", + "attachment": false, + "reasoning": false, + "tool_call": true, + "temperature": true, + "knowledge": "2025-04", + "release_date": "2026-05-20", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 2.0, + "output": 12.0, + "cache_read": 0.2, + "cache_write": 2.5 + }, + "limit": { + "context": 256000, + "output": 64000 + } + }, + { + "id": "hyper/qwen3.6-plus", + "name": "Qwen3.6 Plus", + "family": "qwen", + "attachment": true, + "reasoning": false, + "tool_call": true, + "temperature": true, + "knowledge": "2025-04", + "release_date": "2026-05-20", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 2.0, + "output": 6.0, + "cache_read": 0.2, + "cache_write": 2.5 + }, + "limit": { + "context": 1000000, + "output": 64000 + } + }, + { + "id": "hyper/qwen3.7-max", + "name": "Qwen3.7 Max", + "family": "qwen", + "attachment": false, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2026-05-28", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 2.5, + "output": 7.5, + "cache_read": 0.5 + }, + "limit": { + "context": 1000000, + "output": 64000 + } + }, + { + "id": "hyper/qwen3.7-plus", + "name": "Qwen3.7 Plus", + "family": "qwen", + "attachment": true, + "reasoning": false, + "tool_call": true, + "temperature": true, + "knowledge": "2025-04", + "release_date": "2026-06-15", + "last_updated": "2026-07-22", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.2, + "output": 4.8, + "cache_read": 0.24 + }, + "limit": { + "context": 1000000, + "output": 64000 + } + }, { "id": "iflowcn/deepseek-r1", "name": "DeepSeek-R1", @@ -59312,6 +59586,37 @@ "output": 128000 } }, + { + "id": "kenari/claude-fable-5", + "name": "Claude Fable 5", + "family": "claude-fable", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-01-31", + "release_date": "2026-06-09", + "last_updated": "2026-06-09", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1000000, + "output": 128000 + } + }, { "id": "kenari/claude-opus-4.7", "name": "Claude Opus 4.7", @@ -59375,16 +59680,16 @@ } }, { - "id": "kenari/claude-sonnet-4.6", - "name": "Claude Sonnet 4.6", + "id": "kenari/claude-sonnet-5", + "name": "Claude Sonnet 5", "family": "claude-sonnet", "attachment": true, "reasoning": true, "tool_call": true, - "temperature": true, - "knowledge": "2025-08-31", - "release_date": "2026-02-17", - "last_updated": "2026-03-13", + "temperature": false, + "knowledge": "2026-01-31", + "release_date": "2026-06-30", + "last_updated": "2026-06-30", "modalities": { "input": [ "text", @@ -59402,7 +59707,7 @@ }, "limit": { "context": 1000000, - "output": 64000 + "output": 128000 } }, { @@ -59493,32 +59798,102 @@ } }, { - "id": "kenari/deepseek-v4-pro:free", - "name": "DeepSeek V4 Pro (Free)", - "family": "deepseek-thinking", - "attachment": false, + "id": "kenari/gemini-2.5-flash", + "name": "Gemini 2.5 Flash", + "family": "gemini-flash", + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, - "knowledge": "2025-05", - "release_date": "2026-04-24", - "last_updated": "2026-04-24", + "knowledge": "2025-01", + "release_date": "2025-06-17", + "last_updated": "2025-06-17", "modalities": { "input": [ - "text" + "text", + "image", + "audio", + "video", + "pdf" ], "output": [ "text" ] }, - "open_weights": true, + "open_weights": false, "cost": { "input": 0.0, "output": 0.0 }, "limit": { - "context": 1000000, - "output": 384000 + "context": 1048576, + "output": 65536 + } + }, + { + "id": "kenari/gemini-2.5-flash-lite", + "name": "Gemini 2.5 Flash-Lite", + "family": "gemini-flash-lite", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2025-06-17", + "last_updated": "2025-06-17", + "modalities": { + "input": [ + "text", + "image", + "audio", + "video", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1048576, + "output": 65536 + } + }, + { + "id": "kenari/gemini-3.1-flash-lite", + "name": "Gemini 3.1 Flash Lite", + "family": "gemini-flash-lite", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-05-07", + "last_updated": "2026-05-07", + "modalities": { + "input": [ + "text", + "image", + "video", + "audio", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1048576, + "output": 65536 } }, { @@ -59550,6 +59925,35 @@ "output": 32768 } }, + { + "id": "kenari/glm-4.7-flash:free", + "name": "GLM-4.7-Flash (Free)", + "family": "glm-flash", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-04", + "release_date": "2026-01-19", + "last_updated": "2026-01-19", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 200000, + "output": 131072 + } + }, { "id": "kenari/glm-5.1", "name": "GLM-5.1", @@ -59667,6 +60071,99 @@ "output": 128000 } }, + { + "id": "kenari/gpt-5.6-luna", + "name": "GPT-5.6 Luna", + "family": "gpt-luna", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-02-16", + "release_date": "2026-07-09", + "last_updated": "2026-07-09", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1050000, + "output": 128000 + } + }, + { + "id": "kenari/gpt-5.6-sol", + "name": "GPT-5.6 Sol", + "family": "gpt-sol", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-02-16", + "release_date": "2026-07-09", + "last_updated": "2026-07-09", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1050000, + "output": 128000 + } + }, + { + "id": "kenari/gpt-5.6-terra", + "name": "GPT-5.6 Terra", + "family": "gpt-terra", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-02-16", + "release_date": "2026-07-09", + "last_updated": "2026-07-09", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1050000, + "output": 128000 + } + }, { "id": "kenari/gpt-image-2", "name": "GPT-Image-2", @@ -59752,20 +60249,19 @@ } }, { - "id": "kenari/grok-4.3", - "name": "Grok 4.3", + "id": "kenari/grok-4.5", + "name": "Grok 4.5", "family": "grok", "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, - "release_date": "2026-04-17", - "last_updated": "2026-04-17", + "release_date": "2026-07-08", + "last_updated": "2026-07-08", "modalities": { "input": [ "text", - "image", - "pdf" + "image" ], "output": [ "text" @@ -59777,8 +60273,8 @@ "output": 0.0 }, "limit": { - "context": 1000000, - "output": 30000 + "context": 500000, + "output": 500000 } }, { @@ -59842,6 +60338,37 @@ "output": 262144 } }, + { + "id": "kenari/kimi-k2-6:free", + "name": "Kimi K2.6 (Free)", + "family": "kimi-k2", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-04-21", + "last_updated": "2026-04-21", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, { "id": "kenari/kimi-k2-7-code", "name": "Kimi K2.7 Code", @@ -59873,6 +60400,67 @@ "output": 262144 } }, + { + "id": "kenari/kimi-k2-7-code:free", + "name": "Kimi K2.7 Code (Free)", + "family": "kimi-k2", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2025-01", + "release_date": "2026-06-12", + "last_updated": "2026-06-12", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, + { + "id": "kenari/kimi-k3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1048576, + "output": 131072 + } + }, { "id": "kenari/mimo-v2-5", "name": "MiMo-V2.5", @@ -59934,6 +60522,38 @@ "output": 131072 } }, + { + "id": "kenari/mimo-v2-5:free", + "name": "MiMo-V2.5 (Free)", + "family": "mimo", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2024-12", + "release_date": "2026-04-22", + "last_updated": "2026-04-22", + "modalities": { + "input": [ + "text", + "image", + "audio", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1048576, + "output": 131072 + } + }, { "id": "kenari/minimax-m3", "name": "MiniMax-M3", @@ -59964,11 +60584,123 @@ "output": 128000 } }, + { + "id": "kenari/nemotron-3-nano-30b-a3b", + "name": "Nemotron 3 Nano 30B A3B", + "family": "nemotron", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-12-15", + "last_updated": "2025-12-15", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, + { + "id": "kenari/nemotron-3-super-120b-a12b", + "name": "Nemotron 3 Super 120B A12B", + "family": "nemotron", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-03-11", + "last_updated": "2026-03-11", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, + { + "id": "kenari/nemotron-3-super-120b-a12b:free", + "name": "Nemotron 3 Super 120B A12B (Free)", + "family": "nemotron", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-03-11", + "last_updated": "2026-03-11", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, + { + "id": "kenari/nemotron-3-ultra-550b-a55b", + "name": "Nemotron 3 Ultra 550B A55B", + "family": "nemotron", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-06-04", + "last_updated": "2026-06-04", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1000000, + "output": 128000 + } + }, { "id": "kenari/qwen3-7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -59978,7 +60710,8 @@ "modalities": { "input": [ "text", - "image" + "image", + "video" ], "output": [ "text" @@ -69509,6 +70242,37 @@ "output": 131072 } }, + { + "id": "kimi-for-coding/k3-256k", + "name": "Kimi K3-256K", + "family": "kimi-k3", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0, + "cache_read": 0.0, + "cache_write": 0.0 + }, + "limit": { + "context": 262144, + "output": 131072 + } + }, { "id": "kimi-for-coding/kimi-for-coding", "name": "Kimi K2.7 Code", @@ -70208,6 +70972,34 @@ "output": 16384 } }, + { + "id": "llmgateway/cosmos3-super-reasoner", + "name": "Cosmos 3 Super Reasoner", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-06-01", + "last_updated": "2026-06-01", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.1, + "output": 0.3 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, { "id": "llmgateway/custom", "name": "Custom Model", @@ -70288,9 +71080,9 @@ }, "open_weights": true, "cost": { - "input": 0.14, - "output": 0.28, - "cache_read": 0.0028 + "input": 0.076, + "output": 0.153, + "cache_read": 0.014 }, "limit": { "context": 1050000, @@ -70725,6 +71517,35 @@ "output": 65536 } }, + { + "id": "llmgateway/gemma-3-27b", + "name": "Gemma 3 27B", + "family": "gemma", + "attachment": true, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2025-03-12", + "last_updated": "2025-03-12", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.1, + "output": 0.3 + }, + "limit": { + "context": 110000, + "output": 110000 + } + }, { "id": "llmgateway/gemma-4-26b-a4b-it", "name": "Gemma 4 26B A4B IT", @@ -70775,8 +71596,9 @@ }, "open_weights": true, "cost": { - "input": 0.13, - "output": 0.38 + "input": 0.102, + "output": 0.297, + "cache_read": 0.012 }, "limit": { "context": 262144, @@ -71229,13 +72051,13 @@ }, "open_weights": true, "cost": { - "input": 1.26, - "output": 3.96, - "cache_read": 0.234, + "input": 0.8, + "output": 2.55, + "cache_read": 0.16, "cache_write": 0.0 }, "limit": { - "context": 1024000, + "context": 1048576, "output": 131072 } }, @@ -71516,6 +72338,35 @@ "output": 16384 } }, + { + "id": "llmgateway/gpt-4o-mini-transcribe", + "name": "GPT-4o Mini Transcribe", + "family": "gpt", + "attachment": true, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2025-03-20", + "last_updated": "2025-03-20", + "modalities": { + "input": [ + "text", + "audio" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.25, + "output": 5.0 + }, + "limit": { + "context": 16000, + "output": 16000 + } + }, { "id": "llmgateway/gpt-4o-search-preview", "name": "GPT-4o Search Preview", @@ -71545,6 +72396,35 @@ "output": 16384 } }, + { + "id": "llmgateway/gpt-4o-transcribe", + "name": "GPT-4o Transcribe", + "family": "gpt", + "attachment": true, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2025-03-20", + "last_updated": "2025-03-20", + "modalities": { + "input": [ + "text", + "audio" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 2.5, + "output": 10.0 + }, + "limit": { + "context": 16000, + "output": 16000 + } + }, { "id": "llmgateway/gpt-5", "name": "GPT-5", @@ -72285,8 +73165,9 @@ }, "open_weights": false, "cost": { - "input": 0.05, - "output": 0.25 + "input": 0.032, + "output": 0.14, + "cache_read": 0.032 }, "limit": { "context": 131072, @@ -72627,6 +73508,62 @@ "output": 256000 } }, + { + "id": "llmgateway/hermes-4-405b", + "name": "Hermes 4 405B", + "family": "hermes", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-08-26", + "last_updated": "2025-08-26", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.0, + "output": 3.0 + }, + "limit": { + "context": 131072, + "output": 131072 + } + }, + { + "id": "llmgateway/hermes-4-70b", + "name": "Hermes 4 70B", + "family": "hermes", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-08-26", + "last_updated": "2025-08-26", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.13, + "output": 0.4 + }, + "limit": { + "context": 131072, + "output": 131072 + } + }, { "id": "llmgateway/kimi-k2", "name": "Kimi K2", @@ -72690,7 +73627,7 @@ "id": "llmgateway/kimi-k2.5", "name": "Kimi K2.5", "family": "kimi-k2", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": false, @@ -72741,9 +73678,9 @@ }, "open_weights": true, "cost": { - "input": 0.4, - "output": 2.2, - "cache_read": 0.08 + "input": 0.22, + "output": 1.137, + "cache_read": 0.048 }, "limit": { "context": 262144, @@ -73135,6 +74072,34 @@ "output": 131072 } }, + { + "id": "llmgateway/minicpm-v-4.5", + "name": "MiniCPM-V 4.5", + "attachment": true, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2025-08-26", + "last_updated": "2025-08-26", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.658, + "output": 1.11 + }, + "limit": { + "context": 32000, + "output": 32000 + } + }, { "id": "llmgateway/minimax-m2", "name": "MiniMax-M2", @@ -73300,9 +74265,9 @@ }, "open_weights": true, "cost": { - "input": 0.3, - "output": 1.2, - "cache_read": 0.06, + "input": 0.08, + "output": 0.32, + "cache_read": 0.017, "cache_write": 0.375 }, "limit": { @@ -73367,7 +74332,7 @@ "cache_read": 0.06 }, "limit": { - "context": 524288, + "context": 1048576, "output": 128000 } }, @@ -73578,6 +74543,91 @@ "output": 32000 } }, + { + "id": "llmgateway/nemotron-3-nano-30b", + "name": "Nemotron 3 Nano 30B", + "family": "nemotron", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-12-15", + "last_updated": "2025-12-15", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.06, + "output": 0.24 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, + { + "id": "llmgateway/nemotron-3-nano-omni", + "name": "Nemotron 3 Nano Omni", + "family": "nemotron", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-04-28", + "last_updated": "2026-04-28", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.06, + "output": 0.24 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, + { + "id": "llmgateway/nemotron-3-super-120b", + "name": "Nemotron 3 Super 120B", + "family": "nemotron", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-03-11", + "last_updated": "2026-03-11", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.3, + "output": 0.9 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, { "id": "llmgateway/nemotron-3-ultra-550b", "name": "Nemotron 3 Ultra 550B A55B", @@ -73603,7 +74653,7 @@ "cache_read": 0.15 }, "limit": { - "context": 262144, + "context": 1048576, "output": 128000 } }, @@ -73935,11 +74985,11 @@ }, "open_weights": true, "cost": { - "input": 0.13, - "output": 0.4 + "input": 0.25, + "output": 0.75 }, "limit": { - "context": 32768, + "context": 32000, "output": 8192 } }, @@ -74019,8 +75069,8 @@ }, "open_weights": true, "cost": { - "input": 0.2, - "output": 0.6 + "input": 0.3, + "output": 3.0 }, "limit": { "context": 262000, @@ -74080,7 +75130,7 @@ "output": 0.3 }, "limit": { - "context": 32768, + "context": 40960, "output": 16384 } }, @@ -74596,7 +75646,7 @@ "id": "llmgateway/qwen3.6-plus", "name": "Qwen3.6 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -74659,7 +75709,7 @@ "id": "llmgateway/qwen3.7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -74669,7 +75719,8 @@ "modalities": { "input": [ "text", - "image" + "image", + "video" ], "output": [ "text" @@ -75905,7 +76956,7 @@ "id": "merge-gateway/alibaba/qwen3.6-plus", "name": "Qwen3.6 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -77428,7 +78479,7 @@ "id": "merge-gateway/moonshotai/kimi-k2.5", "name": "Kimi K2.5", "family": "kimi-k2", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": false, @@ -100186,6 +101237,34 @@ "output": 8000 } }, + { + "id": "nebius/moonshotai/Kimi-K3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 3.0, + "output": 15.0 + }, + "limit": { + "context": 1048576, + "output": 8000 + } + }, { "id": "nebius/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1", "name": "Llama-3.1-Nemotron-Ultra-253B-v1", @@ -105247,7 +106326,7 @@ } }, { - "id": "nvidia/abacusai/dracarys-llama-3_1-70b-instruct", + "id": "nvidia/abacusai/dracarys-llama-3.1-70b-instruct", "name": "dracarys-llama-3.1-70b-instruct", "attachment": false, "reasoning": false, @@ -105531,6 +106610,64 @@ "output": 4096 } }, + { + "id": "nvidia/google/gemma-3-12b-it", + "name": "Gemma 3 12B IT", + "family": "gemma", + "attachment": true, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2025-03-12", + "last_updated": "2025-03-12", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 131072, + "output": 16384 + } + }, + { + "id": "nvidia/google/gemma-3-4b-it", + "name": "Gemma 3 4B IT", + "family": "gemma", + "attachment": true, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2025-03-12", + "last_updated": "2025-03-12", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 131072, + "output": 16384 + } + }, { "id": "nvidia/google/gemma-3n-e2b-it", "name": "Gemma 3n E2b It", @@ -106098,7 +107235,36 @@ } }, { - "id": "nvidia/mistralai/mistral-7b-instruct-v03", + "id": "nvidia/mistralai/ministral-14b-instruct", + "name": "Ministral 3 14B Instruct 2512", + "family": "ministral", + "attachment": true, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2025-12-02", + "last_updated": "2025-12-02", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 262144, + "output": 16384 + } + }, + { + "id": "nvidia/mistralai/mistral-7b-instruct-v0.3", "name": "Mistral-7B-Instruct-v0.3", "attachment": false, "reasoning": false, @@ -106182,6 +107348,35 @@ "output": 32768 } }, + { + "id": "nvidia/mistralai/mistral-medium-3.5-128b", + "name": "Mistral Medium 3.5", + "family": "mistral-medium", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-04-29", + "last_updated": "2026-04-29", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 262144, + "output": 32768 + } + }, { "id": "nvidia/mistralai/mistral-nemotron", "name": "mistral-nemotron", @@ -106435,6 +107630,35 @@ "output": 4096 } }, + { + "id": "nvidia/nvidia/cosmos-reason2-8b", + "name": "Cosmos Reason2 8B", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-12-01", + "last_updated": "2025-12-01", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 131072, + "output": 16384 + } + }, { "id": "nvidia/nvidia/cosmos-transfer1-7b", "name": "cosmos-transfer1-7b", @@ -106521,7 +107745,92 @@ } }, { - "id": "nvidia/nvidia/llama-3_1-nemotron-safety-guard-8b-v3", + "id": "nvidia/nvidia/llama-3.1-nemotron-70b-instruct", + "name": "Llama 3.1 Nemotron 70B Instruct", + "family": "nemotron", + "attachment": false, + "reasoning": false, + "tool_call": true, + "temperature": true, + "release_date": "2025-04-15", + "last_updated": "2025-04-15", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 128000, + "output": 8192 + } + }, + { + "id": "nvidia/nvidia/llama-3.1-nemotron-nano-8b-v1", + "name": "Llama 3.1 Nemotron Nano 8B v1", + "family": "nemotron", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-03-18", + "last_updated": "2025-03-18", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 131072, + "output": 16384 + } + }, + { + "id": "nvidia/nvidia/llama-3.1-nemotron-nano-vl-8b-v1", + "name": "Llama 3.1 Nemotron Nano VL 8B v1", + "family": "nemotron", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-04-10", + "last_updated": "2025-04-10", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 32768, + "output": 16384 + } + }, + { + "id": "nvidia/nvidia/llama-3.1-nemotron-safety-guard-8b-v3", "name": "llama-3.1-nemotron-safety-guard-8b-v3", "family": "nemotron", "attachment": false, @@ -106548,6 +107857,90 @@ "output": 4096 } }, + { + "id": "nvidia/nvidia/llama-3.1-nemotron-ultra-253b-v1", + "name": "Llama 3.1 Nemotron Ultra 253B", + "family": "nemotron", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-04-07", + "last_updated": "2025-04-07", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 128000, + "output": 16384 + } + }, + { + "id": "nvidia/nvidia/llama-3.3-nemotron-super-49b-v1", + "name": "Llama 3.3 Nemotron Super 49B v1", + "family": "nemotron", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-04-07", + "last_updated": "2025-04-07", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 131072, + "output": 65536 + } + }, + { + "id": "nvidia/nvidia/llama-3.3-nemotron-super-49b-v1.5", + "name": "Llama 3.3 Nemotron Super 49B v1.5", + "family": "nemotron", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-07-25", + "last_updated": "2025-07-25", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 131072, + "output": 65536 + } + }, { "id": "nvidia/nvidia/llama-3_2-nemoretriever-300m-embed-v1", "name": "llama-3_2-nemoretriever-300m-embed-v1", @@ -106863,6 +108256,36 @@ "output": 8192 } }, + { + "id": "nvidia/nvidia/nemotron-nano-12b-v2-vl", + "name": "Nemotron Nano 12B v2 VL", + "family": "nemotron", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2025-10-28", + "last_updated": "2025-10-28", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 128000, + "output": 128000 + } + }, { "id": "nvidia/nvidia/nemotron-voicechat", "name": "nemotron-voicechat", @@ -107003,7 +108426,7 @@ } }, { - "id": "nvidia/nvidia/riva-translate-4b-instruct-v1_1", + "id": "nvidia/nvidia/riva-translate-4b-instruct-v1.1", "name": "riva-translate-4b-instruct-v1_1", "attachment": false, "reasoning": false, @@ -107277,6 +108700,34 @@ "output": 4096 } }, + { + "id": "nvidia/poolside/laguna-xs-2.1", + "name": "Laguna XS 2.1", + "family": "laguna", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-02", + "last_updated": "2026-07-02", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 262144, + "output": 16384 + } + }, { "id": "nvidia/qwen/qwen-image", "name": "Qwen Image", @@ -107564,7 +109015,37 @@ } }, { - "id": "nvidia/upstage/solar-10_7b-instruct", + "id": "nvidia/thinkingmachines/inkling", + "name": "Inkling", + "family": "ling", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-15", + "last_updated": "2026-07-15", + "modalities": { + "input": [ + "text", + "image", + "audio" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 1048576, + "output": 16384 + } + }, + { + "id": "nvidia/upstage/solar-10.7b-instruct", "name": "solar-10.7b-instruct", "attachment": false, "reasoning": false, @@ -107618,6 +109099,422 @@ "output": 131072 } }, + { + "id": "ofox/anthropic/claude-fable-5", + "name": "Claude Fable 5", + "family": "claude-fable", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-01-31", + "release_date": "2026-06-09", + "last_updated": "2026-06-09", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 10.0, + "output": 50.0, + "cache_read": 1.0, + "cache_write": 12.5 + }, + "limit": { + "context": 1000000, + "output": 128000 + } + }, + { + "id": "ofox/anthropic/claude-opus-4.8", + "name": "Claude Opus 4.8", + "family": "claude-opus", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-01", + "release_date": "2026-05-28", + "last_updated": "2026-05-28", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 5.0, + "output": 25.0, + "cache_read": 0.5, + "cache_write": 6.25 + }, + "limit": { + "context": 1000000, + "output": 128000 + } + }, + { + "id": "ofox/anthropic/claude-sonnet-5", + "name": "Claude Sonnet 5", + "family": "claude-sonnet", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-01-31", + "release_date": "2026-06-30", + "last_updated": "2026-06-30", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 2.0, + "output": 10.0, + "cache_read": 0.2, + "cache_write": 2.5 + }, + "limit": { + "context": 1000000, + "output": 128000 + } + }, + { + "id": "ofox/bailian/qwen3.7-max", + "name": "Qwen3.7 Max", + "family": "qwen", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-05-21", + "last_updated": "2026-05-21", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 2.5, + "output": 7.5, + "cache_read": 0.5, + "cache_write": 3.125 + }, + "limit": { + "context": 1000000, + "output": 65536 + } + }, + { + "id": "ofox/deepseek/deepseek-v4-pro", + "name": "DeepSeek V4 Pro", + "family": "deepseek-thinking", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-05", + "release_date": "2026-04-24", + "last_updated": "2026-04-24", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.45, + "output": 0.88, + "cache_read": 0.0037 + }, + "limit": { + "context": 1000000, + "output": 384000 + } + }, + { + "id": "ofox/google/gemini-3.1-pro-preview", + "name": "Gemini 3.1 Pro Preview", + "family": "gemini-pro", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-02-19", + "last_updated": "2026-02-19", + "modalities": { + "input": [ + "text", + "image", + "video", + "audio", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 2.0, + "output": 12.0, + "cache_read": 0.2 + }, + "limit": { + "context": 1048576, + "output": 65536 + } + }, + { + "id": "ofox/moonshotai/kimi-k2.6", + "name": "Kimi K2.6", + "family": "kimi-k2", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-04-21", + "last_updated": "2026-04-21", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.95, + "output": 4.0, + "cache_read": 0.16 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, + { + "id": "ofox/openai/gpt-5.5", + "name": "GPT-5.5", + "family": "gpt", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2025-12-01", + "release_date": "2026-04-23", + "last_updated": "2026-04-23", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 5.0, + "output": 30.0, + "cache_read": 0.5 + }, + "limit": { + "context": 1050000, + "output": 128000 + } + }, + { + "id": "ofox/openai/gpt-5.6-luna", + "name": "GPT-5.6 Luna", + "family": "gpt-luna", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-02-16", + "release_date": "2026-07-09", + "last_updated": "2026-07-09", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.0, + "output": 6.0, + "cache_read": 0.1, + "cache_write": 1.25 + }, + "limit": { + "context": 1050000, + "output": 128000 + } + }, + { + "id": "ofox/openai/gpt-5.6-sol", + "name": "GPT-5.6 Sol", + "family": "gpt-sol", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-02-16", + "release_date": "2026-07-09", + "last_updated": "2026-07-09", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 5.0, + "output": 30.0, + "cache_read": 0.5, + "cache_write": 6.25 + }, + "limit": { + "context": 1050000, + "output": 128000 + } + }, + { + "id": "ofox/openai/gpt-5.6-terra", + "name": "GPT-5.6 Terra", + "family": "gpt-terra", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-02-16", + "release_date": "2026-07-09", + "last_updated": "2026-07-09", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 2.5, + "output": 15.0, + "cache_read": 0.25, + "cache_write": 3.125 + }, + "limit": { + "context": 1050000, + "output": 128000 + } + }, + { + "id": "ofox/x-ai/grok-4.3", + "name": "Grok 4.3", + "family": "grok", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-04-17", + "last_updated": "2026-04-17", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.25, + "output": 2.5, + "cache_read": 0.2 + }, + "limit": { + "context": 1000000, + "output": 30000 + } + }, + { + "id": "ofox/z-ai/glm-5.2", + "name": "GLM-5.2", + "family": "glm", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-06-13", + "last_updated": "2026-06-13", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 1.4, + "output": 4.4, + "cache_read": 0.26 + }, + "limit": { + "context": 1000000, + "output": 131072 + } + }, { "id": "ollama-cloud/deepseek-v4-flash", "name": "deepseek-v4-flash", @@ -107866,6 +109763,32 @@ "output": 262144 } }, + { + "id": "ollama-cloud/kimi-k3", + "name": "kimi-k3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-27", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": {}, + "limit": { + "context": 1048576, + "output": 131072 + } + }, { "id": "ollama-cloud/minimax-m2.5", "name": "minimax-m2.5", @@ -108374,68 +110297,6 @@ "output": 128000 } }, - { - "id": "openai/gpt-5-chat", - "name": "GPT-5 Chat (latest)", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": false, - "temperature": true, - "knowledge": "2024-09-30", - "release_date": "2025-08-07", - "last_updated": "2025-08-07", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.25, - "output": 10.0, - "cache_read": 0.125 - }, - "limit": { - "context": 400000, - "output": 128000 - } - }, - { - "id": "openai/gpt-5-codex", - "name": "GPT-5-Codex", - "family": "gpt-codex", - "attachment": false, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2024-09-30", - "release_date": "2025-09-15", - "last_updated": "2025-09-15", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.25, - "output": 10.0, - "cache_read": 0.125 - }, - "limit": { - "context": 400000, - "output": 128000 - } - }, { "id": "openai/gpt-5-mini", "name": "GPT-5 Mini", @@ -108559,130 +110420,6 @@ "output": 128000 } }, - { - "id": "openai/gpt-5.1-chat", - "name": "GPT-5.1 Chat", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2024-09-30", - "release_date": "2025-11-13", - "last_updated": "2025-11-13", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.25, - "output": 10.0, - "cache_read": 0.125 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, - { - "id": "openai/gpt-5.1-codex", - "name": "GPT-5.1 Codex", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2024-09-30", - "release_date": "2025-11-13", - "last_updated": "2025-11-13", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.25, - "output": 10.0, - "cache_read": 0.125 - }, - "limit": { - "context": 400000, - "output": 128000 - } - }, - { - "id": "openai/gpt-5.1-codex-max", - "name": "GPT-5.1 Codex Max", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2024-09-30", - "release_date": "2025-11-13", - "last_updated": "2025-11-13", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.25, - "output": 10.0, - "cache_read": 0.125 - }, - "limit": { - "context": 400000, - "output": 128000 - } - }, - { - "id": "openai/gpt-5.1-codex-mini", - "name": "GPT-5.1 Codex mini", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2024-09-30", - "release_date": "2025-11-13", - "last_updated": "2025-11-13", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 0.25, - "output": 2.0, - "cache_read": 0.025 - }, - "limit": { - "context": 400000, - "output": 128000 - } - }, { "id": "openai/gpt-5.2", "name": "GPT-5.2", @@ -108745,38 +110482,6 @@ "output": 16384 } }, - { - "id": "openai/gpt-5.2-codex", - "name": "GPT-5.2 Codex", - "family": "gpt-codex", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2025-08-31", - "release_date": "2025-12-11", - "last_updated": "2025-12-11", - "modalities": { - "input": [ - "text", - "image", - "pdf" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.75, - "output": 14.0, - "cache_read": 0.175 - }, - "limit": { - "context": 400000, - "output": 128000 - } - }, { "id": "openai/gpt-5.2-pro", "name": "GPT-5.2 Pro", @@ -109458,37 +111163,6 @@ "output": 100000 } }, - { - "id": "openai/o3-deep-research", - "name": "o3-deep-research", - "family": "o", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2024-05", - "release_date": "2024-06-26", - "last_updated": "2024-06-26", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 10.0, - "output": 40.0, - "cache_read": 2.5 - }, - "limit": { - "context": 200000, - "output": 100000 - } - }, { "id": "openai/o3-mini", "name": "o3-mini", @@ -109580,37 +111254,6 @@ "output": 100000 } }, - { - "id": "openai/o4-mini-deep-research", - "name": "o4-mini-deep-research", - "family": "o-mini", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2024-05", - "release_date": "2024-06-26", - "last_updated": "2024-06-26", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 2.0, - "output": 8.0, - "cache_read": 0.5 - }, - "limit": { - "context": 200000, - "output": 100000 - } - }, { "id": "openai/text-embedding-3-large", "name": "text-embedding-3-large", @@ -109877,6 +111520,35 @@ "output": 500000 } }, + { + "id": "opencode-go/hy3", + "name": "Hy3", + "family": "Hy", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-06", + "last_updated": "2026-07-06", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.14, + "output": 0.58, + "cache_read": 0.035 + }, + "limit": { + "context": 256000, + "output": 64000 + } + }, { "id": "opencode-go/kimi-k2.5", "name": "Kimi K2.5", @@ -112283,6 +113955,37 @@ "output": 262144 } }, + { + "id": "opencode/kimi-k3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 3.0, + "output": 15.0, + "cache_read": 0.3 + }, + "limit": { + "context": 1048576, + "output": 131072 + } + }, { "id": "opencode/laguna-s-2.1-free", "name": "Laguna S 2.1 Free", @@ -112341,6 +114044,35 @@ "output": 32800 } }, + { + "id": "opencode/ling-3.0-flash-free", + "name": "Ling-3.0-flash Free", + "family": "ling", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-23", + "last_updated": "2026-07-23", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0, + "cache_read": 0.0 + }, + "limit": { + "context": 262144, + "output": 32768 + } + }, { "id": "opencode/mimo-v2-flash-free", "name": "MiMo V2 Flash Free", @@ -113882,13 +115614,13 @@ }, "open_weights": true, "cost": { - "input": 0.25, - "output": 0.8, + "input": 0.22, + "output": 0.85, "cache_read": 0.06 }, "limit": { "context": 262144, - "output": 80000 + "output": 262144 } }, { @@ -114557,13 +116289,13 @@ }, "open_weights": true, "cost": { - "input": 0.0938, - "output": 0.1876, - "cache_read": 0.01876 + "input": 0.14, + "output": 0.28, + "cache_read": 0.028 }, "limit": { "context": 1048576, - "output": 65536 + "output": 393216 } }, { @@ -114962,7 +116694,7 @@ "output": 3.0 }, "limit": { - "context": 131072, + "context": 65536, "output": 65536 } }, @@ -115003,23 +116735,23 @@ }, { "id": "openrouter/google/gemini-3.1-flash-lite-image", - "name": "Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)", - "family": "gemini", + "name": "Nano Banana 2 Lite", + "family": "gemini-flash-lite", "attachment": true, "reasoning": true, "tool_call": false, "temperature": true, - "knowledge": "2025-01-01", + "knowledge": "2025-01", "release_date": "2026-06-30", "last_updated": "2026-06-30", "modalities": { "input": [ - "image", - "text" + "text", + "image" ], "output": [ - "image", - "text" + "text", + "image" ] }, "open_weights": false, @@ -115029,7 +116761,7 @@ }, "limit": { "context": 65536, - "output": 65536 + "output": 66000 } }, { @@ -115323,8 +117055,9 @@ }, "open_weights": true, "cost": { - "input": 0.1, - "output": 0.3 + "input": 0.08, + "output": 0.45, + "cache_read": 0.04 }, "limit": { "context": 262144, @@ -115412,12 +117145,13 @@ }, "open_weights": true, "cost": { - "input": 0.07, - "output": 0.34 + "input": 0.14, + "output": 0.42, + "cache_read": 0.05 }, "limit": { "context": 262144, - "output": 16384 + "output": 262144 } }, { @@ -115472,12 +117206,12 @@ }, "open_weights": true, "cost": { - "input": 0.12, - "output": 0.37 + "input": 0.14, + "output": 0.4 }, "limit": { "context": 262144, - "output": 16384 + "output": 262144 } }, { @@ -115518,8 +117252,8 @@ "reasoning": false, "tool_call": false, "temperature": true, - "release_date": "2026-03-30", - "last_updated": "2026-03-30", + "release_date": "2026-03-25", + "last_updated": "2026-03-25", "modalities": { "input": [ "text", @@ -115548,8 +117282,8 @@ "reasoning": false, "tool_call": false, "temperature": true, - "release_date": "2026-03-30", - "last_updated": "2026-03-30", + "release_date": "2026-03-25", + "last_updated": "2026-03-25", "modalities": { "input": [ "text", @@ -115742,6 +117476,34 @@ "output": 32768 } }, + { + "id": "openrouter/inclusionai/ling-3.0-flash:free", + "name": "Ling-3.0-flash (free)", + "family": "ling", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-23", + "last_updated": "2026-07-23", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 262144, + "output": 32768 + } + }, { "id": "openrouter/inclusionai/ring-2.6-1t", "name": "Ring-2.6-1T", @@ -115771,62 +117533,6 @@ "output": 65536 } }, - { - "id": "openrouter/inflection/inflection-3-pi", - "name": "Inflection 3 Pi", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2024-10-31", - "release_date": "2024-10-11", - "last_updated": "2024-10-11", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 2.5, - "output": 10.0 - }, - "limit": { - "context": 8000, - "output": 1024 - } - }, - { - "id": "openrouter/inflection/inflection-3-productivity", - "name": "Inflection 3 Productivity", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": true, - "knowledge": "2024-10-31", - "release_date": "2024-10-11", - "last_updated": "2024-10-11", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 2.5, - "output": 10.0 - }, - "limit": { - "context": 8000, - "output": 1024 - } - }, { "id": "openrouter/kwaipilot/kat-coder-air-v2.5", "name": "KAT-Coder-Air V2.5", @@ -116081,12 +117787,12 @@ }, "open_weights": true, "cost": { - "input": 0.0509, - "output": 0.335 + "input": 0.05, + "output": 0.33 }, "limit": { "context": 131072, - "output": 80000 + "output": 131072 } }, { @@ -116377,9 +118083,8 @@ }, "open_weights": true, "cost": { - "input": 0.3, - "output": 1.2, - "cache_read": 0.03 + "input": 0.255, + "output": 1.02 }, "limit": { "context": 204800, @@ -117165,9 +118870,9 @@ }, "open_weights": true, "cost": { - "input": 0.684, - "output": 3.42, - "cache_read": 0.144 + "input": 0.646, + "output": 2.72, + "cache_read": 0.1088 }, "limit": { "context": 262144, @@ -117196,9 +118901,9 @@ }, "open_weights": true, "cost": { - "input": 0.82, - "output": 3.75, - "cache_read": 0.16 + "input": 0.73, + "output": 3.5, + "cache_read": 0.15 }, "limit": { "context": 262144, @@ -117212,7 +118917,7 @@ "attachment": true, "reasoning": true, "tool_call": true, - "temperature": false, + "temperature": true, "release_date": "2026-07-16", "last_updated": "2026-07-16", "modalities": { @@ -117574,8 +119279,8 @@ }, "open_weights": true, "cost": { - "input": 0.08, - "output": 0.45 + "input": 0.085, + "output": 0.4 }, "limit": { "context": 1000000, @@ -118089,64 +119794,6 @@ "output": 16384 } }, - { - "id": "openrouter/openai/gpt-4o-mini-search-preview", - "name": "GPT-4o-mini Search Preview", - "family": "o-mini", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": false, - "knowledge": "2023-10-31", - "release_date": "2025-03-12", - "last_updated": "2025-03-12", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 0.15, - "output": 0.6 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, - { - "id": "openrouter/openai/gpt-4o-search-preview", - "name": "GPT-4o Search Preview", - "family": "gpt", - "attachment": false, - "reasoning": false, - "tool_call": false, - "temperature": false, - "knowledge": "2023-10-31", - "release_date": "2025-03-12", - "last_updated": "2025-03-12", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 2.5, - "output": 10.0 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, { "id": "openrouter/openai/gpt-5", "name": "GPT-5", @@ -118179,38 +119826,6 @@ "output": 128000 } }, - { - "id": "openrouter/openai/gpt-5-chat", - "name": "GPT-5 Chat", - "family": "gpt-codex", - "attachment": true, - "reasoning": false, - "tool_call": false, - "temperature": false, - "knowledge": "2024-09-30", - "release_date": "2025-08-07", - "last_updated": "2025-08-07", - "modalities": { - "input": [ - "pdf", - "image", - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.25, - "output": 10.0, - "cache_read": 0.125 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, { "id": "openrouter/openai/gpt-5-codex", "name": "GPT-5-Codex", @@ -118459,11 +120074,11 @@ "cost": { "input": 1.25, "output": 10.0, - "cache_read": 0.125 + "cache_read": 0.13 }, "limit": { "context": 128000, - "output": 16384 + "output": 32000 } }, { @@ -118490,7 +120105,7 @@ "cost": { "input": 1.25, "output": 10.0, - "cache_read": 0.125 + "cache_read": 0.13 }, "limit": { "context": 400000, @@ -118552,11 +120167,11 @@ "cost": { "input": 0.25, "output": 2.0, - "cache_read": 0.025 + "cache_read": 0.03 }, "limit": { "context": 400000, - "output": 100000 + "output": 128000 } }, { @@ -118993,10 +120608,10 @@ }, "open_weights": false, "cost": { - "input": 1.0, - "output": 6.0, - "cache_read": 0.1, - "cache_write": 1.25 + "input": 0.5, + "output": 3.0, + "cache_read": 0.05, + "cache_write": 0.625 }, "limit": { "context": 1050000, @@ -119026,10 +120641,10 @@ }, "open_weights": false, "cost": { - "input": 1.0, - "output": 6.0, - "cache_read": 0.1, - "cache_write": 1.25 + "input": 0.5, + "output": 3.0, + "cache_read": 0.05, + "cache_write": 0.625 }, "limit": { "context": 1050000, @@ -119125,10 +120740,10 @@ }, "open_weights": false, "cost": { - "input": 2.5, - "output": 15.0, - "cache_read": 0.25, - "cache_write": 3.125 + "input": 1.25, + "output": 7.5, + "cache_read": 0.125, + "cache_write": 1.5625 }, "limit": { "context": 1050000, @@ -119158,10 +120773,10 @@ }, "open_weights": false, "cost": { - "input": 2.5, - "output": 15.0, - "cache_read": 0.25, - "cache_write": 3.125 + "input": 1.25, + "output": 7.5, + "cache_read": 0.125, + "cache_write": 1.5625 }, "limit": { "context": 1050000, @@ -119308,8 +120923,7 @@ "open_weights": true, "cost": { "input": 0.03, - "output": 0.13, - "cache_read": 0.03 + "output": 0.14 }, "limit": { "context": 131072, @@ -120363,12 +121977,12 @@ }, "open_weights": true, "cost": { - "input": 0.12, - "output": 0.24 + "input": 0.2275, + "output": 0.91 }, "limit": { "context": 131072, - "output": 16384 + "output": 8192 } }, { @@ -120450,12 +122064,12 @@ }, "open_weights": true, "cost": { - "input": 0.13, - "output": 0.52 + "input": 0.12, + "output": 0.5 }, "limit": { "context": 131072, - "output": 8192 + "output": 16384 } }, { @@ -120479,8 +122093,8 @@ }, "open_weights": true, "cost": { - "input": 0.1, - "output": 0.3 + "input": 0.04815, + "output": 0.19305 }, "limit": { "context": 262144, @@ -120925,12 +122539,12 @@ }, "open_weights": true, "cost": { - "input": 0.13, - "output": 0.52 + "input": 0.15, + "output": 0.6 }, "limit": { "context": 262144, - "output": 32768 + "output": 16384 } }, { @@ -121102,12 +122716,12 @@ }, "open_weights": true, "cost": { - "input": 0.26, - "output": 2.6 + "input": 0.195, + "output": 1.56 }, "limit": { "context": 262144, - "output": 81920 + "output": 65536 } }, { @@ -121132,8 +122746,9 @@ }, "open_weights": true, "cost": { - "input": 0.14, - "output": 1.0 + "input": 0.15, + "output": 1.0, + "cache_read": 0.05 }, "limit": { "context": 262144, @@ -121314,8 +122929,9 @@ }, "open_weights": true, "cost": { - "input": 0.45, - "output": 2.7 + "input": 0.3, + "output": 2.0, + "cache_read": 0.15 }, "limit": { "context": 262144, @@ -121445,6 +123061,38 @@ "output": 65536 } }, + { + "id": "openrouter/qwen/qwen3.7-flash", + "name": "Qwen3.7 Flash", + "family": "qwen", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-27", + "last_updated": "2026-07-27", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.03, + "output": 0.13, + "cache_read": 0.006, + "cache_write": 0.038 + }, + "limit": { + "context": 1000000, + "output": 65536 + } + }, { "id": "openrouter/qwen/qwen3.7-max", "name": "Qwen3.7 Max", @@ -121828,7 +123476,7 @@ { "id": "openrouter/tencent/hy3", "name": "Hy3", - "family": "hy3", + "family": "Hy", "attachment": false, "reasoning": true, "tool_call": true, @@ -121845,13 +123493,13 @@ }, "open_weights": true, "cost": { - "input": 0.14, - "output": 0.58, - "cache_read": 0.035 + "input": 0.132, + "output": 0.528, + "cache_read": 0.033 }, "limit": { "context": 262144, - "output": 262144 + "output": 128000 } }, { @@ -122024,7 +123672,7 @@ "cache_read": 0.17 }, "limit": { - "context": 524288, + "context": 1048576, "output": 1048576 } }, @@ -122537,12 +124185,13 @@ }, "open_weights": true, "cost": { - "input": 0.0605, - "output": 0.4 + "input": 0.06, + "output": 0.4, + "cache_read": 0.01 }, "limit": { "context": 202752, - "output": 131072 + "output": 16384 } }, { @@ -122652,9 +124301,9 @@ }, "open_weights": true, "cost": { - "input": 0.8204, - "output": 2.5784, - "cache_read": 0.15236 + "input": 0.7378, + "output": 2.3188, + "cache_read": 0.13702 }, "limit": { "context": 1048576, @@ -122898,7 +124547,7 @@ "attachment": true, "reasoning": true, "tool_call": true, - "temperature": false, + "temperature": true, "release_date": "2026-04-27", "last_updated": "2026-04-27", "modalities": { @@ -122918,7 +124567,7 @@ }, "limit": { "context": 1048576, - "output": 262144 + "output": 1048576 } }, { @@ -123867,7 +125516,7 @@ "id": "orcarouter/kimi/kimi-k2.5", "name": "Kimi K2.5", "family": "kimi-k2", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": false, @@ -125257,7 +126906,7 @@ "id": "orcarouter/qwen/qwen3.6-plus", "name": "Qwen3.6 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -128734,7 +130383,7 @@ "id": "pioneer/qwen3.6-plus", "name": "Qwen3.6 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -128797,7 +130446,7 @@ "id": "pioneer/qwen3.7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -128807,7 +130456,8 @@ "modalities": { "input": [ "text", - "image" + "image", + "video" ], "output": [ "text" @@ -132794,15 +134444,15 @@ } }, { - "id": "poolside/poolside/laguna-xs-2.1", - "name": "Laguna XS 2.1", + "id": "poolside/poolside/laguna-s-2.1", + "name": "Laguna S 2.1", "family": "laguna", "attachment": false, "reasoning": true, "tool_call": true, "temperature": true, - "release_date": "2026-07-02", - "last_updated": "2026-07-02", + "release_date": "2026-07-21", + "last_updated": "2026-07-21", "modalities": { "input": [ "text" @@ -132819,20 +134469,20 @@ "cache_write": 0.0 }, "limit": { - "context": 262144, + "context": 1048576, "output": 32768 } }, { - "id": "poolside/poolside/laguna-xs.2", - "name": "Laguna XS.2", + "id": "poolside/poolside/laguna-xs-2.1", + "name": "Laguna XS 2.1", "family": "laguna", "attachment": false, "reasoning": true, "tool_call": true, "temperature": true, - "release_date": "2026-04-28", - "last_updated": "2026-06-13", + "release_date": "2026-07-02", + "last_updated": "2026-07-02", "modalities": { "input": [ "text" @@ -138376,6 +140026,32 @@ "output": 65536 } }, + { + "id": "sap-ai-core/gemini-embedding", + "name": "Gemini Embedding 001", + "family": "gemini", + "attachment": false, + "reasoning": false, + "tool_call": false, + "temperature": false, + "knowledge": "2025-05", + "release_date": "2025-05-20", + "last_updated": "2025-05-20", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": {}, + "limit": { + "context": 2048, + "output": 1 + } + }, { "id": "sap-ai-core/gpt-4.1", "name": "gpt-4.1", @@ -143925,6 +145601,36 @@ "output": 65536 } }, + { + "id": "synthetic/hf:moonshotai/Kimi-K3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-16", + "last_updated": "2026-07-27", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 3.0, + "output": 15.0, + "cache_read": 0.45 + }, + "limit": { + "context": 524288, + "output": 65536 + } + }, { "id": "synthetic/hf:nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4", "name": "Nemotron 3 Super 120B A12B", @@ -144595,7 +146301,7 @@ } }, { - "id": "thinkingmachines/inkling", + "id": "thinkingmachines/thinkingmachines/Inkling", "name": "Inkling", "family": "ling", "attachment": true, @@ -144607,9 +146313,37 @@ "modalities": { "input": [ "text", - "image", - "audio", - "pdf" + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 1.87, + "output": 4.68, + "cache_read": 0.374 + }, + "limit": { + "context": 65536, + "output": 65536 + } + }, + { + "id": "thinkingmachines/thinkingmachines/Inkling:peft:262144", + "name": "Inkling (256K)", + "family": "ling", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-15", + "last_updated": "2026-07-15", + "modalities": { + "input": [ + "text", + "image" ], "output": [ "text" @@ -144622,8 +146356,8 @@ "cache_read": 0.748 }, "limit": { - "context": 256000, - "output": 256000 + "context": 262144, + "output": 262144 } }, { @@ -145552,6 +147286,37 @@ "output": 131072 } }, + { + "id": "togetherai/moonshotai/Kimi-K3", + "name": "Kimi K3", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-16", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 3.0, + "output": 15.0, + "cache_read": 0.3 + }, + "limit": { + "context": 1048576, + "output": 131072 + } + }, { "id": "togetherai/nvidia/nemotron-3-ultra-550b-a55b", "name": "Nemotron 3 Ultra 550B A55B", @@ -147349,38 +149114,6 @@ "output": 128000 } }, - { - "id": "venice/claude-opus-4.7-fast", - "name": "Claude Opus 4.7 Fast", - "family": "claude-opus", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": false, - "knowledge": "2026-01-31", - "release_date": "2026-05-14", - "last_updated": "2026-06-11", - "modalities": { - "input": [ - "text", - "image" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 36.0, - "output": 180.0, - "cache_read": 3.6, - "cache_write": 45.0 - }, - "limit": { - "context": 1000000, - "output": 128000 - } - }, { "id": "venice/claude-opus-4.8", "name": "Claude Opus 4.8", @@ -148297,11 +150030,11 @@ "text" ] }, - "open_weights": false, + "open_weights": true, "cost": { - "input": 3.75, - "output": 18.75, - "cache_read": 0.375 + "input": 4.6875, + "output": 23.4375, + "cache_read": 0.46875 }, "limit": { "context": 1000000, @@ -148592,34 +150325,6 @@ "output": 32768 } }, - { - "id": "venice/nvidia-nemotron-cascade-2-30b-a3b", - "name": "Nemotron Cascade 2 30B A3B", - "family": "nemotron", - "attachment": false, - "reasoning": true, - "tool_call": true, - "temperature": true, - "release_date": "2026-03-24", - "last_updated": "2026-06-11", - "modalities": { - "input": [ - "text" - ], - "output": [ - "text" - ] - }, - "open_weights": true, - "cost": { - "input": 0.14, - "output": 0.8 - }, - "limit": { - "context": 256000, - "output": 32768 - } - }, { "id": "venice/olafangensan-glm-4.7-flash-heretic", "name": "GLM 4.7 Flash Heretic", @@ -149450,7 +151155,7 @@ "id": "venice/qwen3-6-35b-a3b", "name": "Qwen 3.6 35B A3B", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -149458,7 +151163,8 @@ "last_updated": "2026-07-22", "modalities": { "input": [ - "text" + "text", + "image" ], "output": [ "text" @@ -149466,7 +151172,7 @@ }, "open_weights": true, "cost": { - "input": 0.15, + "input": 0.1, "output": 1.0, "cache_read": 0.05 }, @@ -149558,6 +151264,36 @@ "output": 16384 } }, + { + "id": "venice/seed-2.1-turbo", + "name": "Seed 2.1 Turbo", + "family": "seed", + "attachment": true, + "reasoning": true, + "tool_call": true, + "release_date": "2026-06-28", + "last_updated": "2026-07-24", + "modalities": { + "input": [ + "text", + "image", + "video" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.625, + "output": 3.125, + "cache_read": 0.125 + }, + "limit": { + "context": 256000, + "output": 65536 + } + }, { "id": "venice/venice-uncensored-1.2", "name": "Venice Uncensored 1.2", @@ -150637,10 +152373,10 @@ }, "open_weights": false, "cost": { - "input": 1.25, - "output": 3.75, - "cache_read": 0.25, - "cache_write": 1.5625 + "input": 2.5, + "output": 7.5, + "cache_read": 0.5, + "cache_write": 3.125 }, "limit": { "context": 991000, @@ -151294,39 +153030,6 @@ "output": 128000 } }, - { - "id": "vercel/anthropic/claude-opus-4.7-fast", - "name": "Claude Opus 4.7 (Fast)", - "family": "claude-opus", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": true, - "knowledge": "2026-01-31", - "release_date": "2026-04-16", - "last_updated": "2026-04-16", - "modalities": { - "input": [ - "text", - "image", - "pdf" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 30.0, - "output": 150.0, - "cache_read": 3.0, - "cache_write": 37.5 - }, - "limit": { - "context": 1000000, - "output": 128000 - } - }, { "id": "vercel/anthropic/claude-opus-4.8", "name": "Claude Opus 4.8", @@ -152748,38 +154451,6 @@ "output": 32768 } }, - { - "id": "vercel/google/gemini-3-pro-preview", - "name": "Gemini 3 Pro Preview", - "family": "gemini-pro", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": true, - "knowledge": "2025-01", - "release_date": "2025-11-18", - "last_updated": "2025-11-18", - "modalities": { - "input": [ - "text", - "image", - "pdf" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 2.0, - "output": 12.0, - "cache_read": 0.2 - }, - "limit": { - "context": 1000000, - "output": 64000 - } - }, { "id": "vercel/google/gemini-3.1-flash-image", "name": "Nano Banana 2", @@ -152884,6 +154555,7 @@ "reasoning": true, "tool_call": false, "temperature": true, + "knowledge": "2025-01", "release_date": "2026-06-30", "last_updated": "2026-06-30", "modalities": { @@ -152907,38 +154579,6 @@ "output": 4096 } }, - { - "id": "vercel/google/gemini-3.1-flash-lite-preview", - "name": "Gemini 3.1 Flash Lite Preview", - "family": "gemini", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": true, - "knowledge": "2025-01", - "release_date": "2026-03-03", - "last_updated": "2026-03-03", - "modalities": { - "input": [ - "text", - "image", - "pdf" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 0.25, - "output": 1.5, - "cache_read": 0.03 - }, - "limit": { - "context": 1000000, - "output": 65000 - } - }, { "id": "vercel/google/gemini-3.1-pro-preview", "name": "Gemini 3.1 Pro Preview", @@ -152948,7 +154588,7 @@ "tool_call": true, "temperature": true, "knowledge": "2025-01", - "release_date": "2025-11-18", + "release_date": "2026-02-19", "last_updated": "2026-02-19", "modalities": { "input": [ @@ -153101,8 +154741,9 @@ "reasoning": false, "tool_call": false, "temperature": true, + "knowledge": "2025-11", "release_date": "2026-03-10", - "last_updated": "2026-03-23", + "last_updated": "2026-04-22", "modalities": { "input": [ "text" @@ -153491,6 +155132,33 @@ "output": 16384 } }, + { + "id": "vercel/inclusionai/ling-3.0-flash-free", + "name": "Ling 3.0 Flash", + "family": "ling", + "attachment": false, + "reasoning": true, + "tool_call": true, + "release_date": "2026-07-23", + "last_updated": "2026-07-23", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 256000, + "output": 256000 + } + }, { "id": "vercel/interfaze/interfaze-beta", "name": "Interfaze Beta", @@ -154937,6 +156605,37 @@ "output": 131072 } }, + { + "id": "vercel/moonshotai/kimi-k3-fast", + "name": "Kimi K3 Fast", + "family": "kimi-k3", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "release_date": "2026-07-27", + "last_updated": "2026-07-16", + "modalities": { + "input": [ + "text", + "image", + "pdf" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 4.5, + "output": 22.5, + "cache_read": 0.45 + }, + "limit": { + "context": 1000000, + "output": 131072 + } + }, { "id": "vercel/morph/morph-v3-fast", "name": "Morph v3 Fast", @@ -155472,44 +157171,11 @@ "output": 128000 } }, - { - "id": "vercel/openai/gpt-5-chat", - "name": "GPT-5 Chat", - "family": "gpt", - "attachment": true, - "reasoning": true, - "tool_call": true, - "temperature": true, - "knowledge": "2024-10", - "release_date": "2025-08-07", - "last_updated": "2025-08-07", - "modalities": { - "input": [ - "text", - "image", - "pdf" - ], - "output": [ - "text", - "image" - ] - }, - "open_weights": false, - "cost": { - "input": 1.25, - "output": 10.0, - "cache_read": 0.125 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, { "id": "vercel/openai/gpt-5-codex", "name": "GPT-5-Codex", - "family": "gpt-codex", - "attachment": false, + "family": "gpt", + "attachment": true, "reasoning": true, "tool_call": true, "temperature": false, @@ -155519,7 +157185,8 @@ "modalities": { "input": [ "text", - "image" + "image", + "pdf" ], "output": [ "text" @@ -155822,38 +157489,6 @@ "output": 128000 } }, - { - "id": "vercel/openai/gpt-5.2-chat", - "name": "GPT-5.2 Chat", - "family": "gpt", - "attachment": true, - "reasoning": false, - "tool_call": true, - "temperature": true, - "knowledge": "2024-10", - "release_date": "2025-12-11", - "last_updated": "2025-08-07", - "modalities": { - "input": [ - "text", - "image", - "pdf" - ], - "output": [ - "text" - ] - }, - "open_weights": false, - "cost": { - "input": 1.75, - "output": 14.0, - "cache_read": 0.175 - }, - "limit": { - "context": 128000, - "output": 16384 - } - }, { "id": "vercel/openai/gpt-5.2-codex", "name": "GPT-5.2-Codex", @@ -157041,6 +158676,63 @@ "output": 8000 } }, + { + "id": "vercel/poolside/laguna-s-2.1", + "name": "Laguna S 2.1", + "family": "laguna-s", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-20", + "last_updated": "2026-07-21", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.1, + "output": 0.2, + "cache_read": 0.01 + }, + "limit": { + "context": 1000000, + "output": 131072 + } + }, + { + "id": "vercel/poolside/laguna-s-2.1-free", + "name": "Laguna S 2.1 Free", + "family": "laguna-s", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-21", + "last_updated": "2026-07-21", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.0, + "output": 0.0 + }, + "limit": { + "context": 256000, + "output": 32768 + } + }, { "id": "vercel/prodia/flux-fast-schnell", "name": "Flux Schnell", @@ -157382,6 +159074,35 @@ "output": 256000 } }, + { + "id": "vercel/tencent/hy3", + "name": "Hy3", + "family": "hy3", + "attachment": false, + "reasoning": true, + "tool_call": true, + "temperature": true, + "release_date": "2026-07-06", + "last_updated": "2026-07-06", + "modalities": { + "input": [ + "text" + ], + "output": [ + "text" + ] + }, + "open_weights": true, + "cost": { + "input": 0.14, + "output": 0.58, + "cache_read": 0.035 + }, + "limit": { + "context": 262144, + "output": 262144 + } + }, { "id": "vercel/thinkingmachines/inkling", "name": "Inkling", @@ -158115,7 +159836,7 @@ "tool_call": false, "temperature": true, "release_date": "2026-06-22", - "last_updated": "2026-06-22", + "last_updated": "2026-05-30", "modalities": { "input": [ "text" @@ -158690,7 +160411,7 @@ "tool_call": true, "temperature": true, "release_date": "2026-06-23", - "last_updated": "2026-06-16", + "last_updated": "2026-06-13", "modalities": { "input": [ "text" @@ -159805,9 +161526,9 @@ }, "open_weights": true, "cost": { - "input": 0.29, - "output": 1.2, - "cache_read": 0.06 + "input": 0.23, + "output": 0.96, + "cache_read": 0.05 }, "limit": { "context": 262144, @@ -160192,9 +161913,9 @@ }, "open_weights": true, "cost": { - "input": 0.12, - "output": 0.35, - "cache_read": 0.09 + "input": 0.1, + "output": 0.34, + "cache_read": 0.1 }, "limit": { "context": 262144, @@ -160402,9 +162123,9 @@ }, "open_weights": true, "cost": { - "input": 0.95, - "output": 4.0, - "cache_read": 0.16 + "input": 0.65, + "output": 3.41, + "cache_read": 0.15 }, "limit": { "context": 262144, @@ -160433,9 +162154,9 @@ }, "open_weights": true, "cost": { - "input": 0.94, - "output": 4.0, - "cache_read": 0.19 + "input": 0.71, + "output": 3.5, + "cache_read": 0.15 }, "limit": { "context": 262144, @@ -160520,9 +162241,9 @@ }, "open_weights": true, "cost": { - "input": 0.04, - "output": 0.14, - "cache_read": 0.04 + "input": 0.03, + "output": 0.17, + "cache_read": 0.03 }, "limit": { "context": 131072, @@ -160607,9 +162328,9 @@ }, "open_weights": true, "cost": { - "input": 1.39, - "output": 4.4, - "cache_read": 0.26 + "input": 0.76, + "output": 2.42, + "cache_read": 0.14 }, "limit": { "context": 262144, @@ -160886,6 +162607,32 @@ "output": 0 } }, + { + "id": "x-ai/grok-imagine-video-1.5", + "name": "Grok Imagine Video 1.5", + "family": "grok", + "attachment": true, + "reasoning": false, + "tool_call": false, + "temperature": false, + "release_date": "2026-05-30", + "last_updated": "2026-05-30", + "modalities": { + "input": [ + "image", + "pdf" + ], + "output": [ + "video" + ] + }, + "open_weights": false, + "cost": {}, + "limit": { + "context": 1024, + "output": 0 + } + }, { "id": "xiaomi-token-plan-ams/mimo-v2-pro", "name": "MiMo-V2-Pro", @@ -161699,7 +163446,7 @@ "open_weights": false, "cost": { "input": 3.0, - "output": 18.0, + "output": 18.5, "cache_read": 0.3 }, "limit": { @@ -161707,6 +163454,316 @@ "output": 128000 } }, + { + "id": "xpersona/claude-haiku-4.5", + "name": "Claude Haiku 4.5 (latest)", + "family": "claude-haiku", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-02-28", + "release_date": "2025-10-15", + "last_updated": "2025-10-15", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.6, + "output": 3.7, + "cache_read": 0.06 + }, + "limit": { + "context": 200000, + "output": 128000 + } + }, + { + "id": "xpersona/claude-opus-4.8", + "name": "Claude Opus 4.8", + "family": "claude-opus", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-01", + "release_date": "2026-05-28", + "last_updated": "2026-05-28", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.5, + "output": 9.25, + "cache_read": 0.15 + }, + "limit": { + "context": 200000, + "output": 128000 + } + }, + { + "id": "xpersona/claude-sonnet-4.6", + "name": "Claude Sonnet 4.6", + "family": "claude-sonnet", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-08-31", + "release_date": "2026-02-17", + "last_updated": "2026-03-13", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.9, + "output": 5.55, + "cache_read": 0.09 + }, + "limit": { + "context": 200000, + "output": 128000 + } + }, + { + "id": "xpersona/gemini-3.5-flash", + "name": "Gemini 3.5 Flash", + "family": "gemini-flash", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": true, + "knowledge": "2025-01", + "release_date": "2026-05-19", + "last_updated": "2026-05-19", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.55, + "output": 12.2, + "cache_read": 0.155 + }, + "limit": { + "context": 1000000, + "output": 128000 + } + }, + { + "id": "xpersona/gpt-5.4", + "name": "GPT-5.4", + "family": "gpt", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2025-08-31", + "release_date": "2026-03-05", + "last_updated": "2026-03-05", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.75, + "output": 6.0, + "cache_read": 0.075 + }, + "limit": { + "context": 1050000, + "output": 128000 + } + }, + { + "id": "xpersona/gpt-5.4-mini", + "name": "GPT-5.4 mini", + "family": "gpt-mini", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2025-08-31", + "release_date": "2026-03-17", + "last_updated": "2026-03-17", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 0.375, + "output": 4.0, + "cache_read": 0.0375 + }, + "limit": { + "context": 272000, + "output": 128000 + } + }, + { + "id": "xpersona/gpt-5.5", + "name": "GPT-5.5", + "family": "gpt", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2025-12-01", + "release_date": "2026-04-23", + "last_updated": "2026-04-23", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.5, + "output": 12.0, + "cache_read": 0.15 + }, + "limit": { + "context": 1050000, + "output": 128000 + } + }, + { + "id": "xpersona/gpt-5.6", + "name": "GPT-5.6", + "family": "gpt-sol", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-02-16", + "release_date": "2026-07-09", + "last_updated": "2026-07-09", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.5, + "output": 12.0, + "cache_read": 0.15 + }, + "limit": { + "context": 372000, + "output": 128000 + } + }, + { + "id": "xpersona/gpt-5.6-sol", + "name": "GPT-5.6 Sol", + "family": "gpt-sol", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-02-16", + "release_date": "2026-07-09", + "last_updated": "2026-07-09", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.5, + "output": 12.0, + "cache_read": 0.15 + }, + "limit": { + "context": 372000, + "output": 128000 + } + }, + { + "id": "xpersona/gpt-5.6-terra", + "name": "GPT-5.6 Terra", + "family": "gpt-terra", + "attachment": true, + "reasoning": true, + "tool_call": true, + "temperature": false, + "knowledge": "2026-02-16", + "release_date": "2026-07-09", + "last_updated": "2026-07-09", + "modalities": { + "input": [ + "text", + "image" + ], + "output": [ + "text" + ] + }, + "open_weights": false, + "cost": { + "input": 1.5, + "output": 2.0, + "cache_read": 0.15 + }, + "limit": { + "context": 372000, + "output": 128000 + } + }, { "id": "xpersona/xpersona-frieren-coder", "name": "Xpersona Frieren 1", @@ -164774,7 +166831,7 @@ "id": "zenmux/qwen/qwen3.7-plus", "name": "Qwen3.7 Plus", "family": "qwen", - "attachment": false, + "attachment": true, "reasoning": true, "tool_call": true, "temperature": true, @@ -164784,7 +166841,8 @@ "modalities": { "input": [ "text", - "image" + "image", + "video" ], "output": [ "text" diff --git a/crates/goose-provider-types/src/canonical/data/provider_metadata.json b/crates/goose-provider-types/src/canonical/data/provider_metadata.json index c88043b3c..b394dd0b5 100644 --- a/crates/goose-provider-types/src/canonical/data/provider_metadata.json +++ b/crates/goose-provider-types/src/canonical/data/provider_metadata.json @@ -30,7 +30,7 @@ "env": [ "ABLIT_KEY" ], - "model_count": 1 + "model_count": 2 }, { "id": "ai-router", @@ -43,6 +43,17 @@ ], "model_count": 5 }, + { + "id": "aiand", + "display_name": "ai&", + "npm": "@ai-sdk/openai-compatible", + "api": "https://api.aiand.com/v1", + "doc": "https://docs.aiand.com/", + "env": [ + "AIAND_API_KEY" + ], + "model_count": 8 + }, { "id": "aki-io", "display_name": "AKI.IO", @@ -107,7 +118,7 @@ "env": [ "ALIBABA_TOKEN_PLAN_API_KEY" ], - "model_count": 19 + "model_count": 22 }, { "id": "alibaba-token-plan-cn", @@ -118,7 +129,7 @@ "env": [ "ALIBABA_TOKEN_PLAN_API_KEY" ], - "model_count": 19 + "model_count": 22 }, { "id": "ambient", @@ -184,7 +195,7 @@ "env": [ "BASETEN_API_KEY" ], - "model_count": 14 + "model_count": 16 }, { "id": "berget", @@ -195,7 +206,7 @@ "env": [ "BERGET_API_KEY" ], - "model_count": 8 + "model_count": 9 }, { "id": "blueclaw", @@ -217,7 +228,7 @@ "env": [ "CHUTES_API_KEY" ], - "model_count": 13 + "model_count": 14 }, { "id": "clarifai", @@ -250,7 +261,7 @@ "env": [ "CLINE_API_KEY" ], - "model_count": 10 + "model_count": 11 }, { "id": "cloudferro-sherlock", @@ -284,7 +295,7 @@ "env": [ "CORTECS_API_KEY" ], - "model_count": 56 + "model_count": 57 }, { "id": "crof", @@ -295,7 +306,7 @@ "env": [ "CROF_API_KEY" ], - "model_count": 23 + "model_count": 24 }, { "id": "crossmodel", @@ -306,7 +317,7 @@ "env": [ "CROSSMODEL_API_KEY" ], - "model_count": 44 + "model_count": 46 }, { "id": "daoxe", @@ -362,7 +373,7 @@ "env": [ "DINFERENCE_API_KEY" ], - "model_count": 5 + "model_count": 6 }, { "id": "drun", @@ -395,7 +406,7 @@ "env": [ "EMPIRIOLABS_API_KEY" ], - "model_count": 37 + "model_count": 36 }, { "id": "evroc", @@ -472,7 +483,7 @@ "env": [ "GITHUB_TOKEN" ], - "model_count": 28 + "model_count": 29 }, { "id": "github-models", @@ -507,6 +518,17 @@ ], "model_count": 90 }, + { + "id": "hetzner", + "display_name": "Hetzner", + "npm": "@ai-sdk/openai-compatible", + "api": "https://inference.hetzner.com/api/v1", + "doc": "https://experiments.hetzner.com/docs/inference", + "env": [ + "HETZNER_API_KEY" + ], + "model_count": 1 + }, { "id": "hpc-ai", "display_name": "HPC-AI", @@ -527,7 +549,18 @@ "env": [ "HF_TOKEN" ], - "model_count": 51 + "model_count": 53 + }, + { + "id": "hyper", + "display_name": "Charm Hyper", + "npm": "@ai-sdk/openai-compatible", + "api": "https://hyper.charm.land/v1", + "doc": "https://hyper.charm.land", + "env": [ + "HYPER_API_KEY" + ], + "model_count": 20 }, { "id": "iflowcn", @@ -615,7 +648,7 @@ "env": [ "KENARI_API_KEY" ], - "model_count": 23 + "model_count": 38 }, { "id": "kilo", @@ -637,7 +670,7 @@ "env": [ "KIMI_API_KEY" ], - "model_count": 3 + "model_count": 4 }, { "id": "kuae-cloud-coding-plan", @@ -670,7 +703,7 @@ "env": [ "LLMGATEWAY_API_KEY" ], - "model_count": 174 + "model_count": 185 }, { "id": "llmtr", @@ -912,7 +945,7 @@ "env": [ "NEBIUS_API_KEY" ], - "model_count": 32 + "model_count": 33 }, { "id": "neon", @@ -968,7 +1001,18 @@ "env": [ "NVIDIA_API_KEY" ], - "model_count": 84 + "model_count": 98 + }, + { + "id": "ofox", + "display_name": "Ofox", + "npm": "@ai-sdk/openai-compatible", + "api": "https://api.ofox.ai/v1", + "doc": "https://ofox.ai/docs", + "env": [ + "OFOX_API_KEY" + ], + "model_count": 13 }, { "id": "ollama-cloud", @@ -979,7 +1023,7 @@ "env": [ "OLLAMA_API_KEY" ], - "model_count": 18 + "model_count": 19 }, { "id": "opencode", @@ -990,7 +1034,7 @@ "env": [ "OPENCODE_API_KEY" ], - "model_count": 82 + "model_count": 85 }, { "id": "opencode-go", @@ -1001,7 +1045,7 @@ "env": [ "OPENCODE_API_KEY" ], - "model_count": 21 + "model_count": 22 }, { "id": "orcarouter", @@ -1289,7 +1333,7 @@ "env": [ "SYNTHETIC_API_KEY" ], - "model_count": 7 + "model_count": 8 }, { "id": "tencent-coding-plan", @@ -1338,13 +1382,13 @@ { "id": "thinkingmachines", "display_name": "Thinking Machines", - "npm": "@ai-sdk/openai-compatible", - "api": "https://tinker.thinkingmachines.dev/services/tinker-prod/oai/api/v1", - "doc": "https://tinker-docs.thinkingmachines.ai/tinker/compatible-apis/openai/", + "npm": "@ai-sdk/anthropic", + "api": "https://tinker.thinkingmachines.dev/services/tinker-prod/anthropic/api/v1", + "doc": "https://tinker-docs.thinkingmachines.ai/tinker/compatible-apis/anthropic/", "env": [ "TINKER_API_KEY" ], - "model_count": 1 + "model_count": 2 }, { "id": "tinfoil", @@ -1509,7 +1553,7 @@ "env": [ "XPERSONA_API_KEY" ], - "model_count": 3 + "model_count": 13 }, { "id": "zai", diff --git a/crates/goose-provider-types/src/canonical/name_builder.rs b/crates/goose-provider-types/src/canonical/name_builder.rs index 7dadd1e84..8e1ecefa6 100644 --- a/crates/goose-provider-types/src/canonical/name_builder.rs +++ b/crates/goose-provider-types/src/canonical/name_builder.rs @@ -541,10 +541,6 @@ mod tests { map_to_canonical_model("gcp_vertex_ai", "claude-sonnet-4@20250514", r), Some("google-vertex/claude-sonnet-4".to_string()) ); - assert_eq!( - map_to_canonical_model("gcp_vertex_ai", "claude-3-5-haiku@20241022", r), - Some("google-vertex/claude-3.5-haiku".to_string()) - ); assert_eq!( map_to_canonical_model("gcp_vertex_ai", "claude-sonnet-4-5@20250929", r), Some("google-vertex/claude-sonnet-4.5".to_string()) diff --git a/crates/goose-providers/src/openai.rs b/crates/goose-providers/src/openai.rs index 6b5234108..f19182255 100644 --- a/crates/goose-providers/src/openai.rs +++ b/crates/goose-providers/src/openai.rs @@ -40,22 +40,23 @@ pub const OPEN_AI_DEFAULT_FAST_MODEL: &str = "gpt-4o-mini"; pub const OPEN_AI_KNOWN_MODELS: &[(&str, usize)] = &[ ("gpt-4o", 128_000), ("gpt-4o-mini", 128_000), - ("gpt-4.1", 128_000), - ("gpt-4.1-mini", 128_000), + ("gpt-4.1", 1_047_576), + ("gpt-4.1-mini", 1_047_576), + ("gpt-4.1-nano", 1_047_576), ("o1", 200_000), + ("o1-pro", 200_000), ("o3", 200_000), + ("o3-mini", 200_000), + ("o3-pro", 200_000), ("gpt-3.5-turbo", 16_385), ("gpt-4-turbo", 128_000), - ("o4-mini", 128_000), + ("o4-mini", 200_000), ("gpt-5", 400_000), ("gpt-5-mini", 400_000), ("gpt-5-nano", 400_000), ("gpt-5-pro", 400_000), - ("gpt-5-codex", 400_000), ("gpt-5.1", 400_000), - ("gpt-5.1-codex", 400_000), ("gpt-5.2", 400_000), - ("gpt-5.2-codex", 400_000), ("gpt-5.2-pro", 400_000), ("gpt-5.3-codex", 400_000), ("gpt-5.4", 1_050_000), diff --git a/crates/goose/src/providers/formats/gcpvertexai.rs b/crates/goose/src/providers/formats/gcpvertexai.rs index 8afca2013..f1b011599 100644 --- a/crates/goose/src/providers/formats/gcpvertexai.rs +++ b/crates/goose/src/providers/formats/gcpvertexai.rs @@ -70,17 +70,12 @@ pub const KNOWN_MODELS: &[&str] = &[ "claude-haiku-4-5@20251001", "claude-opus-4@20250514", "claude-sonnet-4@20250514", - "claude-3-5-haiku@20241022", - "claude-3-haiku@20240307", "gemini-3.1-flash-lite", "gemini-3.1-pro-preview", - "gemini-3-pro-preview", "gemini-3-flash-preview", "gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.5-flash-lite", - "gemini-2.0-flash", - "gemini-2.0-flash-lite", ]; /// Represents available GCP Vertex AI models for goose. diff --git a/ui/desktop/package.json b/ui/desktop/package.json index 8937e4a0d..1bcbd767a 100644 --- a/ui/desktop/package.json +++ b/ui/desktop/package.json @@ -1,7 +1,7 @@ { "name": "goose-app", "productName": "Goose", - "version": "1.44.0", + "version": "1.45.0", "description": "Goose App", "engines": { "node": "^24.10.0",