From 7d1abb25a77f220fc3d766f27f56c5eb4a2804bb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Tom=C3=A1s=20Barreiro?= <52393857+BarreiroT@users.noreply.github.com> Date: Tue, 16 Jun 2026 22:28:35 -0300 Subject: [PATCH] Update fixed model list (#11591) --- .../llms/src/catalog/catalog.generated.ts | 821 +++++++++++++----- .../llms/src/providers/builtins.test.ts | 45 - sdk/packages/shared/src/feature-flags.ts | 5 +- 3 files changed, 589 insertions(+), 282 deletions(-) diff --git a/sdk/packages/llms/src/catalog/catalog.generated.ts b/sdk/packages/llms/src/catalog/catalog.generated.ts index 3d2155c82e..d1cc908be2 100644 --- a/sdk/packages/llms/src/catalog/catalog.generated.ts +++ b/sdk/packages/llms/src/catalog/catalog.generated.ts @@ -14,7 +14,7 @@ export const GENERATED_PROVIDER_MODELS: { version: number; providers: Record>; } = { - version: 1781284023269, + version: 1781654469496, providers: { aihubmix: { "glm-5v-turbo": { @@ -1519,6 +1519,51 @@ export const GENERATED_PROVIDER_MODELS: { }, }, baseten: { + "zai-org/GLM-5.2": { + id: "zai-org/GLM-5.2", + name: "GLM-5.2", + contextWindow: 131072, + maxInputTokens: 131072, + maxTokens: 131072, + capabilities: [ + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 1.5, + output: 4.5, + cacheRead: 0.3, + cacheWrite: 0, + }, + releaseDate: "2026-06-13", + family: "glm", + }, + "moonshotai/Kimi-K2.7-Code": { + id: "moonshotai/Kimi-K2.7-Code", + name: "Kimi K2.7 Code", + contextWindow: 262000, + maxInputTokens: 262000, + maxTokens: 262000, + capabilities: [ + "images", + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 0.95, + output: 4, + cacheRead: 0.16, + cacheWrite: 0, + }, + releaseDate: "2026-06-12", + family: "kimi-k2", + }, "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B": { id: "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B", name: "Nemotron Ultra", @@ -3658,6 +3703,76 @@ export const GENERATED_PROVIDER_MODELS: { family: "glm", description: "", }, + "cline-pass/deepseek-v4-flash": { + id: "cline-pass/deepseek-v4-flash", + name: "cline-pass/deepseek-v4-flash", + contextWindow: 1048575, + maxInputTokens: 1048575, + maxTokens: 65536, + capabilities: [ + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 0.098, + output: 0.196, + cacheRead: 0.02, + cacheWrite: 0, + }, + releaseDate: "2026-04-24", + family: "deepseek-flash", + description: "", + }, + "cline-pass/deepseek-v4-pro": { + id: "cline-pass/deepseek-v4-pro", + name: "cline-pass/deepseek-v4-pro", + contextWindow: 1048576, + maxInputTokens: 1048576, + maxTokens: 384000, + capabilities: [ + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 0.435, + output: 0.87, + cacheRead: 0.003625, + cacheWrite: 0, + }, + releaseDate: "2026-04-24", + family: "deepseek-thinking", + description: "", + }, + "cline-pass/kimi-k2.6": { + id: "cline-pass/kimi-k2.6", + name: "cline-pass/kimi-k2.6", + contextWindow: 262142, + maxInputTokens: 262142, + maxTokens: 262142, + capabilities: [ + "images", + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 0.68, + output: 3.41, + cacheRead: 0.34, + cacheWrite: 0, + }, + releaseDate: "2026-04-21", + family: "kimi-k2", + description: "", + }, }, deepseek: { "deepseek-v4-flash": { @@ -3738,6 +3853,90 @@ export const GENERATED_PROVIDER_MODELS: { }, }, fireworks: { + "accounts/fireworks/models/kimi-k2p7-code": { + id: "accounts/fireworks/models/kimi-k2p7-code", + name: "Kimi K2.7 Code", + contextWindow: 262000, + maxInputTokens: 262000, + maxTokens: 262000, + capabilities: [ + "images", + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 0.95, + output: 4, + cacheRead: 0.19, + cacheWrite: 0, + }, + releaseDate: "2026-06-12", + family: "kimi-k2", + }, + "accounts/fireworks/models/minimax-m3": { + id: "accounts/fireworks/models/minimax-m3", + name: "MiniMax-M3", + contextWindow: 512000, + maxInputTokens: 512000, + maxTokens: 512000, + capabilities: ["tools", "reasoning", "temperature", "prompt-cache"], + pricing: { + input: 0.3, + output: 1.2, + cacheRead: 0.06, + cacheWrite: 0, + }, + releaseDate: "2026-06-12", + family: "minimax", + }, + "accounts/fireworks/models/qwen3p7-plus": { + id: "accounts/fireworks/models/qwen3p7-plus", + name: "Qwen 3.7 Plus", + contextWindow: 262144, + maxInputTokens: 262144, + maxTokens: 65536, + capabilities: [ + "images", + "tools", + "reasoning", + "temperature", + "prompt-cache", + ], + pricing: { + input: 0.4, + output: 1.6, + cacheRead: 0.08, + cacheWrite: 0, + }, + releaseDate: "2026-06-12", + family: "qwen", + }, + "accounts/fireworks/routers/kimi-k2p7-code-fast": { + id: "accounts/fireworks/routers/kimi-k2p7-code-fast", + name: "Kimi K2.7 Code Fast", + contextWindow: 262000, + maxInputTokens: 262000, + maxTokens: 262000, + capabilities: [ + "images", + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 2, + output: 8, + cacheRead: 0.38, + cacheWrite: 0, + }, + releaseDate: "2026-06-12", + family: "kimi-k2", + }, "accounts/fireworks/models/deepseek-v4-flash": { id: "accounts/fireworks/models/deepseek-v4-flash", name: "DeepSeek V4 Flash", @@ -3864,28 +4063,6 @@ export const GENERATED_PROVIDER_MODELS: { releaseDate: "2026-04-12", family: "minimax", }, - "accounts/fireworks/models/qwen3p6-plus": { - id: "accounts/fireworks/models/qwen3p6-plus", - name: "Qwen 3.6 Plus", - contextWindow: 262144, - maxInputTokens: 262144, - maxTokens: 65536, - capabilities: [ - "images", - "tools", - "reasoning", - "temperature", - "prompt-cache", - ], - pricing: { - input: 0.5, - output: 3, - cacheRead: 0.1, - cacheWrite: 0, - }, - releaseDate: "2026-04-04", - family: "qwen", - }, "accounts/fireworks/models/glm-5p1": { id: "accounts/fireworks/models/glm-5p1", name: "GLM 5.1", @@ -3918,44 +4095,6 @@ export const GENERATED_PROVIDER_MODELS: { releaseDate: "2026-04-01", family: "glm", }, - "accounts/fireworks/models/minimax-m2p5": { - id: "accounts/fireworks/models/minimax-m2p5", - name: "MiniMax-M2.5", - contextWindow: 196608, - maxInputTokens: 196608, - maxTokens: 196608, - capabilities: ["tools", "reasoning", "temperature", "prompt-cache"], - pricing: { - input: 0.3, - output: 1.2, - cacheRead: 0.03, - cacheWrite: 0, - }, - releaseDate: "2026-02-12", - family: "minimax", - }, - "accounts/fireworks/models/kimi-k2p5": { - id: "accounts/fireworks/models/kimi-k2p5", - name: "Kimi K2.5", - contextWindow: 256000, - maxInputTokens: 256000, - maxTokens: 256000, - capabilities: [ - "images", - "tools", - "reasoning", - "temperature", - "prompt-cache", - ], - pricing: { - input: 0.6, - output: 3, - cacheRead: 0.1, - cacheWrite: 0, - }, - releaseDate: "2026-01-27", - family: "kimi-thinking", - }, "accounts/fireworks/models/gpt-oss-120b": { id: "accounts/fireworks/models/gpt-oss-120b", name: "GPT OSS 120B", @@ -3966,7 +4105,7 @@ export const GENERATED_PROVIDER_MODELS: { pricing: { input: 0.15, output: 0.6, - cacheRead: 0.015, + cacheRead: 0.01, cacheWrite: 0, }, releaseDate: "2025-08-05", @@ -4082,6 +4221,50 @@ export const GENERATED_PROVIDER_MODELS: { releaseDate: "2026-04-02", family: "gemma", }, + "gemma-4-E2B-it": { + id: "gemma-4-E2B-it", + name: "Gemma 4 E2B IT", + contextWindow: 131072, + maxInputTokens: 131072, + maxTokens: 8192, + capabilities: [ + "images", + "tools", + "reasoning", + "structured_output", + "temperature", + ], + pricing: { + input: 0, + output: 0, + cacheRead: 0, + cacheWrite: 0, + }, + releaseDate: "2026-04-02", + family: "gemma", + }, + "gemma-4-E4B-it": { + id: "gemma-4-E4B-it", + name: "Gemma 4 E4B IT", + contextWindow: 131072, + maxInputTokens: 131072, + maxTokens: 8192, + capabilities: [ + "images", + "tools", + "reasoning", + "structured_output", + "temperature", + ], + pricing: { + input: 0, + output: 0, + cacheRead: 0, + cacheWrite: 0, + }, + releaseDate: "2026-04-02", + family: "gemma", + }, "gemini-3.1-pro-preview": { id: "gemini-3.1-pro-preview", name: "Gemini 3.1 Pro Preview", @@ -4202,6 +4385,30 @@ export const GENERATED_PROVIDER_MODELS: { releaseDate: "2025-09-25", family: "gemini-flash-lite", }, + "gemini-2.5-flash": { + id: "gemini-2.5-flash", + name: "Gemini 2.5 Flash", + contextWindow: 1048576, + maxInputTokens: 1048576, + maxTokens: 65536, + capabilities: [ + "images", + "files", + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 0.3, + output: 2.5, + cacheRead: 0.03, + cacheWrite: 0, + }, + releaseDate: "2025-06-17", + family: "gemini-flash", + }, "gemini-2.5-flash-lite": { id: "gemini-2.5-flash-lite", name: "Gemini 2.5 Flash-Lite", @@ -4226,30 +4433,6 @@ export const GENERATED_PROVIDER_MODELS: { releaseDate: "2025-06-17", family: "gemini-flash-lite", }, - "gemini-2.5-flash": { - id: "gemini-2.5-flash", - name: "Gemini 2.5 Flash", - contextWindow: 1048576, - maxInputTokens: 1048576, - maxTokens: 65536, - capabilities: [ - "images", - "files", - "tools", - "reasoning", - "structured_output", - "temperature", - "prompt-cache", - ], - pricing: { - input: 0.3, - output: 2.5, - cacheRead: 0.03, - cacheWrite: 0, - }, - releaseDate: "2025-03-20", - family: "gemini-flash", - }, "gemini-2.5-pro": { id: "gemini-2.5-pro", name: "Gemini 2.5 Pro", @@ -4271,7 +4454,7 @@ export const GENERATED_PROVIDER_MODELS: { cacheRead: 0.125, cacheWrite: 0, }, - releaseDate: "2025-03-20", + releaseDate: "2025-06-17", family: "gemini-pro", }, }, @@ -5015,9 +5198,9 @@ export const GENERATED_PROVIDER_MODELS: { "poolside/laguna-m.1:free": { id: "poolside/laguna-m.1:free", name: "Poolside: Laguna M.1 (free)", - contextWindow: 131072, - maxInputTokens: 131072, - maxTokens: 8192, + contextWindow: 262144, + maxInputTokens: 262144, + maxTokens: 32768, capabilities: ["tools", "reasoning", "temperature"], pricing: { input: 0, @@ -5030,9 +5213,9 @@ export const GENERATED_PROVIDER_MODELS: { "poolside/laguna-xs.2:free": { id: "poolside/laguna-xs.2:free", name: "Poolside: Laguna XS.2 (free)", - contextWindow: 131072, - maxInputTokens: 131072, - maxTokens: 8192, + contextWindow: 262144, + maxInputTokens: 262144, + maxTokens: 32768, capabilities: ["tools", "reasoning", "temperature"], pricing: { input: 0, @@ -8298,6 +8481,21 @@ export const GENERATED_PROVIDER_MODELS: { }, releaseDate: "2024-12-03", }, + "cohere/command-r7b-12-2024": { + id: "cohere/command-r7b-12-2024", + name: "Cohere: Command R7B (12-2024)", + contextWindow: 128000, + maxInputTokens: 128000, + maxTokens: 4000, + capabilities: ["tools", "temperature"], + pricing: { + input: 0.0375, + output: 0.15, + cacheRead: 0, + cacheWrite: 0, + }, + releaseDate: "2024-12-02", + }, "amazon/nova-2-lite-v1": { id: "amazon/nova-2-lite-v1", name: "Amazon: Nova 2 Lite", @@ -8847,21 +9045,6 @@ export const GENERATED_PROVIDER_MODELS: { }, releaseDate: "2024-03-07", }, - "cohere/command-r7b-12-2024": { - id: "cohere/command-r7b-12-2024", - name: "Cohere: Command R7B (12-2024)", - contextWindow: 128000, - maxInputTokens: 128000, - maxTokens: 4000, - capabilities: ["tools", "temperature"], - pricing: { - input: 0.0375, - output: 0.15, - cacheRead: 0, - cacheWrite: 0, - }, - releaseDate: "2024-02-27", - }, "openai/gpt-4-turbo-preview": { id: "openai/gpt-4-turbo-preview", name: "OpenAI: GPT-4 Turbo Preview", @@ -9191,6 +9374,28 @@ export const GENERATED_PROVIDER_MODELS: { releaseDate: "2026-06-12", family: "kimi-k2", }, + "kimi-k2.7-code-highspeed": { + id: "kimi-k2.7-code-highspeed", + name: "Kimi K2.7 Code HighSpeed", + contextWindow: 262144, + maxInputTokens: 262144, + maxTokens: 262144, + capabilities: [ + "images", + "tools", + "reasoning", + "structured_output", + "prompt-cache", + ], + pricing: { + input: 1.9, + output: 8, + cacheRead: 0.38, + cacheWrite: 0, + }, + releaseDate: "2026-06-12", + family: "kimi-k2", + }, "kimi-k2.6": { id: "kimi-k2.6", name: "Kimi K2.6", @@ -9671,6 +9876,22 @@ export const GENERATED_PROVIDER_MODELS: { }, }, ollama: { + "kimi-k2.7-code": { + id: "kimi-k2.7-code", + name: "kimi-k2.7-code", + contextWindow: 262144, + maxInputTokens: 262144, + maxTokens: 262144, + capabilities: ["images", "tools", "reasoning", "structured_output"], + pricing: { + input: 0, + output: 0, + cacheRead: 0, + cacheWrite: 0, + }, + releaseDate: "2026-06-12", + family: "kimi-k2", + }, "nemotron-3-ultra": { id: "nemotron-3-ultra", name: "nemotron-3-ultra", @@ -11007,32 +11228,48 @@ export const GENERATED_PROVIDER_MODELS: { }, }, openrouter: { - "~anthropic/claude-fable-latest": { - id: "~anthropic/claude-fable-latest", - name: "Claude Fable Latest", - contextWindow: 1000000, - maxInputTokens: 1000000, - maxTokens: 128000, + "z-ai/glm-5.2": { + id: "z-ai/glm-5.2", + name: "GLM-5.2", + contextWindow: 1048576, + maxInputTokens: 1048576, + maxTokens: 131072, + capabilities: ["tools", "reasoning", "temperature", "prompt-cache"], + pricing: { + input: 1.4, + output: 4.4, + cacheRead: 0.26, + cacheWrite: 0, + }, + releaseDate: "2026-06-13", + family: "glm", + }, + "moonshotai/kimi-k2.7-code": { + id: "moonshotai/kimi-k2.7-code", + name: "Kimi K2.7 Code", + contextWindow: 262144, + maxInputTokens: 262144, + maxTokens: 262144, capabilities: [ "images", - "files", "tools", "reasoning", "structured_output", + "temperature", "prompt-cache", ], pricing: { - input: 10, - output: 50, - cacheRead: 1, - cacheWrite: 12.5, + input: 0.75, + output: 3.5, + cacheRead: 0.16, + cacheWrite: 0, }, - releaseDate: "2026-06-09", - family: "claude-fable", + releaseDate: "2026-06-12", + family: "kimi-k2", }, - "anthropic/claude-fable-5": { - id: "anthropic/claude-fable-5", - name: "Claude Fable 5", + "~anthropic/claude-fable-latest": { + id: "~anthropic/claude-fable-latest", + name: "Claude Fable Latest", contextWindow: 1000000, maxInputTokens: 1000000, maxTokens: 128000, @@ -11145,6 +11382,7 @@ export const GENERATED_PROVIDER_MODELS: { "images", "tools", "reasoning", + "structured_output", "temperature", "prompt-cache", ], @@ -11561,9 +11799,9 @@ export const GENERATED_PROVIDER_MODELS: { "~moonshotai/kimi-latest": { id: "~moonshotai/kimi-latest", name: "MoonshotAI Kimi Latest", - contextWindow: 262144, - maxInputTokens: 262144, - maxTokens: 262144, + contextWindow: 262142, + maxInputTokens: 262142, + maxTokens: 262142, capabilities: [ "images", "tools", @@ -11573,9 +11811,9 @@ export const GENERATED_PROVIDER_MODELS: { "prompt-cache", ], pricing: { - input: 0.67, - output: 3.39, - cacheRead: 0.14, + input: 0.68, + output: 3.41, + cacheRead: 0.34, cacheWrite: 0, }, releaseDate: "2026-04-27", @@ -11676,9 +11914,9 @@ export const GENERATED_PROVIDER_MODELS: { "deepseek/deepseek-v4-flash": { id: "deepseek/deepseek-v4-flash", name: "DeepSeek V4 Flash", - contextWindow: 1048576, - maxInputTokens: 1048576, - maxTokens: 131072, + contextWindow: 1048575, + maxInputTokens: 1048575, + maxTokens: 65536, capabilities: [ "tools", "reasoning", @@ -11687,9 +11925,9 @@ export const GENERATED_PROVIDER_MODELS: { "prompt-cache", ], pricing: { - input: 0.0983, - output: 0.1966, - cacheRead: 0.0197, + input: 0.098, + output: 0.196, + cacheRead: 0.02, cacheWrite: 0, }, releaseDate: "2026-04-24", @@ -11786,9 +12024,9 @@ export const GENERATED_PROVIDER_MODELS: { "qwen/qwen3.6-27b": { id: "qwen/qwen3.6-27b", name: "Qwen3.6 27B", - contextWindow: 131072, - maxInputTokens: 131072, - maxTokens: 131072, + contextWindow: 262140, + maxInputTokens: 262140, + maxTokens: 262140, capabilities: [ "images", "tools", @@ -11797,8 +12035,8 @@ export const GENERATED_PROVIDER_MODELS: { "temperature", ], pricing: { - input: 0.289, - output: 2.4, + input: 0.2885, + output: 3.17, cacheRead: 0, cacheWrite: 0, }, @@ -11815,6 +12053,7 @@ export const GENERATED_PROVIDER_MODELS: { "images", "tools", "reasoning", + "structured_output", "temperature", "prompt-cache", ], @@ -11896,9 +12135,9 @@ export const GENERATED_PROVIDER_MODELS: { "moonshotai/kimi-k2.6": { id: "moonshotai/kimi-k2.6", name: "Kimi K2.6", - contextWindow: 262144, - maxInputTokens: 262144, - maxTokens: 262144, + contextWindow: 262142, + maxInputTokens: 262142, + maxTokens: 262142, capabilities: [ "images", "tools", @@ -11908,9 +12147,9 @@ export const GENERATED_PROVIDER_MODELS: { "prompt-cache", ], pricing: { - input: 0.67, - output: 3.39, - cacheRead: 0.14, + input: 0.68, + output: 3.41, + cacheRead: 0.34, cacheWrite: 0, }, releaseDate: "2026-04-21", @@ -11946,9 +12185,9 @@ export const GENERATED_PROVIDER_MODELS: { maxTokens: 262144, capabilities: ["tools", "reasoning", "temperature", "prompt-cache"], pricing: { - input: 0.063, - output: 0.21, - cacheRead: 0.021, + input: 0.066, + output: 0.26, + cacheRead: 0.029, cacheWrite: 0, }, releaseDate: "2026-04-20", @@ -12674,7 +12913,7 @@ export const GENERATED_PROVIDER_MODELS: { name: "Qwen3.5 35B-A3B", contextWindow: 262144, maxInputTokens: 262144, - maxTokens: 262144, + maxTokens: 81920, capabilities: [ "images", "tools", @@ -12789,8 +13028,8 @@ export const GENERATED_PROVIDER_MODELS: { "qwen/qwen3.5-397b-a17b": { id: "qwen/qwen3.5-397b-a17b", name: "Qwen3.5 397B-A17B", - contextWindow: 262144, - maxInputTokens: 262144, + contextWindow: 131072, + maxInputTokens: 131072, maxTokens: 65536, capabilities: [ "images", @@ -12800,8 +13039,8 @@ export const GENERATED_PROVIDER_MODELS: { "temperature", ], pricing: { - input: 0.39, - output: 2.34, + input: 0.385, + output: 2.45, cacheRead: 0, cacheWrite: 0, }, @@ -13056,8 +13295,8 @@ export const GENERATED_PROVIDER_MODELS: { "moonshotai/kimi-k2.5": { id: "moonshotai/kimi-k2.5", name: "Kimi K2.5", - contextWindow: 262144, - maxInputTokens: 262144, + contextWindow: 256000, + maxInputTokens: 256000, maxTokens: 262144, capabilities: [ "images", @@ -13065,12 +13304,11 @@ export const GENERATED_PROVIDER_MODELS: { "reasoning", "structured_output", "temperature", - "prompt-cache", ], pricing: { - input: 0.4, - output: 1.9, - cacheRead: 0.09, + input: 0.375, + output: 2.025, + cacheRead: 0, cacheWrite: 0, }, releaseDate: "2026-01", @@ -13210,7 +13448,12 @@ export const GENERATED_PROVIDER_MODELS: { contextWindow: 262144, maxInputTokens: 262144, maxTokens: 228000, - capabilities: ["tools", "reasoning", "temperature"], + capabilities: [ + "tools", + "reasoning", + "structured_output", + "temperature", + ], pricing: { input: 0.05, output: 0.2, @@ -14504,7 +14747,12 @@ export const GENERATED_PROVIDER_MODELS: { contextWindow: 131072, maxInputTokens: 131072, maxTokens: 8192, - capabilities: ["tools", "reasoning", "temperature"], + capabilities: [ + "tools", + "reasoning", + "structured_output", + "temperature", + ], pricing: { input: 0, output: 0, @@ -14590,13 +14838,7 @@ export const GENERATED_PROVIDER_MODELS: { contextWindow: 131070, maxInputTokens: 131070, maxTokens: 131070, - capabilities: [ - "tools", - "reasoning", - "structured_output", - "temperature", - "prompt-cache", - ], + capabilities: ["tools", "reasoning", "temperature", "prompt-cache"], pricing: { input: 0.125, output: 0.85, @@ -15299,8 +15541,8 @@ export const GENERATED_PROVIDER_MODELS: { "deepseek/deepseek-chat-v3-0324": { id: "deepseek/deepseek-chat-v3-0324", name: "DeepSeek V3 0324", - contextWindow: 32768, - maxInputTokens: 32768, + contextWindow: 163840, + maxInputTokens: 163840, maxTokens: 16384, capabilities: [ "tools", @@ -17423,6 +17665,51 @@ export const GENERATED_PROVIDER_MODELS: { }, }, together: { + "moonshotai/Kimi-K2.7-Code": { + id: "moonshotai/Kimi-K2.7-Code", + name: "Kimi K2.7 Code", + contextWindow: 262144, + maxInputTokens: 262144, + maxTokens: 131072, + capabilities: [ + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 0.95, + output: 4, + cacheRead: 0.19, + cacheWrite: 0, + }, + releaseDate: "2026-06-14", + family: "kimi-k2", + }, + "MiniMaxAI/MiniMax-M3": { + id: "MiniMaxAI/MiniMax-M3", + name: "MiniMax-M3", + contextWindow: 524288, + maxInputTokens: 524288, + maxTokens: 250000, + capabilities: [ + "images", + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 0.3, + output: 1.2, + cacheRead: 0.06, + cacheWrite: 0, + }, + releaseDate: "2026-06-12", + family: "minimax", + }, "nvidia/nemotron-3-ultra-550b-a55b": { id: "nvidia/nemotron-3-ultra-550b-a55b", name: "Nemotron 3 Ultra 550B A55B", @@ -17453,8 +17740,8 @@ export const GENERATED_PROVIDER_MODELS: { maxTokens: 500000, capabilities: ["tools", "temperature"], pricing: { - input: 2.5, - output: 7.5, + input: 1.25, + output: 3.75, cacheRead: 0, cacheWrite: 0, }, @@ -17799,12 +18086,34 @@ export const GENERATED_PROVIDER_MODELS: { }, }, "vercel-ai-gateway": { - "anthropic/claude-fable-5": { - id: "anthropic/claude-fable-5", - name: "Claude Fable 5", + "zai/glm-5.2": { + id: "zai/glm-5.2", + name: "GLM 5.2", contextWindow: 1000000, maxInputTokens: 1000000, maxTokens: 128000, + capabilities: [ + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 1.4, + output: 4.4, + cacheRead: 0.26, + cacheWrite: 0, + }, + releaseDate: "2026-06-16", + family: "glm", + }, + "moonshotai/kimi-k2.7-code-highspeed": { + id: "moonshotai/kimi-k2.7-code-highspeed", + name: "Kimi K2.7 Code High Speed", + contextWindow: 262144, + maxInputTokens: 262144, + maxTokens: 0, capabilities: [ "images", "files", @@ -17814,13 +18123,37 @@ export const GENERATED_PROVIDER_MODELS: { "prompt-cache", ], pricing: { - input: 10, - output: 50, - cacheRead: 1, - cacheWrite: 12.5, + input: 1.9, + output: 8, + cacheRead: 0.38, + cacheWrite: 0, }, - releaseDate: "2026-06-09", - family: "claude", + releaseDate: "2026-06-15", + family: "kimi-k2", + }, + "moonshotai/kimi-k2.7-code": { + id: "moonshotai/kimi-k2.7-code", + name: "Kimi K2.7 Code", + contextWindow: 256000, + maxInputTokens: 256000, + maxTokens: 32768, + capabilities: [ + "images", + "files", + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 0.95, + output: 4, + cacheRead: 0.19, + cacheWrite: 0, + }, + releaseDate: "2026-06-12", + family: "kimi-k2", }, "nvidia/nemotron-3-ultra-550b-a55b": { id: "nvidia/nemotron-3-ultra-550b-a55b", @@ -18375,7 +18708,6 @@ export const GENERATED_PROVIDER_MODELS: { "images", "files", "tools", - "reasoning", "structured_output", "temperature", ], @@ -20090,6 +20422,30 @@ export const GENERATED_PROVIDER_MODELS: { releaseDate: "2025-07-09", family: "grok", }, + "google/gemini-2.5-flash": { + id: "google/gemini-2.5-flash", + name: "Gemini 2.5 Flash", + contextWindow: 1048576, + maxInputTokens: 1048576, + maxTokens: 65536, + capabilities: [ + "images", + "files", + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 0.3, + output: 2.5, + cacheRead: 0.03, + cacheWrite: 0, + }, + releaseDate: "2025-06-17", + family: "gemini-flash", + }, "google/gemini-2.5-flash-lite": { id: "google/gemini-2.5-flash-lite", name: "Gemini 2.5 Flash Lite", @@ -20114,6 +20470,30 @@ export const GENERATED_PROVIDER_MODELS: { releaseDate: "2025-06-17", family: "gemini-flash-lite", }, + "google/gemini-2.5-pro": { + id: "google/gemini-2.5-pro", + name: "Gemini 2.5 Pro", + contextWindow: 1048576, + maxInputTokens: 1048576, + maxTokens: 65536, + capabilities: [ + "images", + "files", + "tools", + "reasoning", + "structured_output", + "temperature", + "prompt-cache", + ], + pricing: { + input: 1.25, + output: 10, + cacheRead: 0.125, + cacheWrite: 0, + }, + releaseDate: "2025-06-17", + family: "gemini-pro", + }, "anthropic/claude-opus-4": { id: "anthropic/claude-opus-4", name: "Claude Opus 4", @@ -20398,7 +20778,7 @@ export const GENERATED_PROVIDER_MODELS: { contextWindow: 262144, maxInputTokens: 262144, maxTokens: 16384, - capabilities: ["tools", "temperature"], + capabilities: ["tools", "reasoning", "temperature"], pricing: { input: 0.22, output: 0.88, @@ -20446,7 +20826,7 @@ export const GENERATED_PROVIDER_MODELS: { contextWindow: 262144, maxInputTokens: 262144, maxTokens: 65536, - capabilities: ["tools", "temperature", "prompt-cache"], + capabilities: ["tools", "reasoning", "temperature", "prompt-cache"], pricing: { input: 1.5, output: 7.5, @@ -20472,54 +20852,6 @@ export const GENERATED_PROVIDER_MODELS: { releaseDate: "2025-04-01", family: "qwen", }, - "google/gemini-2.5-flash": { - id: "google/gemini-2.5-flash", - name: "Gemini 2.5 Flash", - contextWindow: 1048576, - maxInputTokens: 1048576, - maxTokens: 65536, - capabilities: [ - "images", - "files", - "tools", - "reasoning", - "structured_output", - "temperature", - "prompt-cache", - ], - pricing: { - input: 0.3, - output: 2.5, - cacheRead: 0.03, - cacheWrite: 0, - }, - releaseDate: "2025-03-20", - family: "gemini-flash", - }, - "google/gemini-2.5-pro": { - id: "google/gemini-2.5-pro", - name: "Gemini 2.5 Pro", - contextWindow: 1048576, - maxInputTokens: 1048576, - maxTokens: 65536, - capabilities: [ - "images", - "files", - "tools", - "reasoning", - "structured_output", - "temperature", - "prompt-cache", - ], - pricing: { - input: 1.25, - output: 10, - cacheRead: 0.125, - cacheWrite: 0, - }, - releaseDate: "2025-03-20", - family: "gemini-pro", - }, "mistral/magistral-medium": { id: "mistral/magistral-medium", name: "Magistral Medium (latest)", @@ -22468,6 +22800,27 @@ export const GENERATED_PROVIDER_MODELS: { }, }, "zai-coding-plan": { + "glm-5.2": { + id: "glm-5.2", + name: "GLM-5.2", + contextWindow: 1000000, + maxInputTokens: 1000000, + maxTokens: 131072, + capabilities: [ + "tools", + "reasoning", + "structured_output", + "temperature", + ], + pricing: { + input: 0, + output: 0, + cacheRead: 0, + cacheWrite: 0, + }, + releaseDate: "2026-06-13", + family: "glm", + }, "glm-5v-turbo": { id: "glm-5v-turbo", name: "GLM-5V-Turbo", diff --git a/sdk/packages/llms/src/providers/builtins.test.ts b/sdk/packages/llms/src/providers/builtins.test.ts index d26902485e..e458492773 100644 --- a/sdk/packages/llms/src/providers/builtins.test.ts +++ b/sdk/packages/llms/src/providers/builtins.test.ts @@ -132,51 +132,6 @@ describe("built-in provider metadata", () => { ); }); - it("includes Claude Fable 5 in Anthropic, OpenRouter, Vercel AI Gateway, and Cline model lists", async () => { - const anthropicModels = await getModelsForProvider("anthropic"); - const openRouterModels = await getModelsForProvider("openrouter"); - const vercelModels = await getModelsForProvider("vercel-ai-gateway"); - const clineModels = await getModelsForProvider("cline"); - - expect(anthropicModels["claude-fable-5"]).toEqual( - expect.objectContaining({ - name: "Claude Fable 5", - contextWindow: 1_000_000, - maxTokens: 128_000, - capabilities: expect.arrayContaining([ - "tools", - "reasoning", - "prompt-cache", - ]), - pricing: expect.objectContaining({ - input: 10, - output: 50, - cacheRead: 1, - cacheWrite: 12.5, - }), - }), - ); - expect(openRouterModels["anthropic/claude-fable-5"]).toEqual( - expect.objectContaining({ - id: "anthropic/claude-fable-5", - name: "Claude Fable 5", - contextWindow: 1_000_000, - maxTokens: 128_000, - }), - ); - expect(vercelModels["anthropic/claude-fable-5"]).toEqual( - expect.objectContaining({ - id: "anthropic/claude-fable-5", - name: "Claude Fable 5", - contextWindow: 1_000_000, - }), - ); - expect(clineModels["anthropic/claude-fable-5"]).toEqual( - openRouterModels["anthropic/claude-fable-5"], - ); - expect(clineModels["anthropic/claude-opus-4.8"]).toBeDefined(); - }); - it("routes native Z.AI providers through GLM thinking metadata", async () => { for (const providerId of ["zai", "zai-coding-plan"] as const) { await expect(getProvider(providerId)).resolves.toMatchObject({ diff --git a/sdk/packages/shared/src/feature-flags.ts b/sdk/packages/shared/src/feature-flags.ts index fe42df6da3..447d5f799e 100644 --- a/sdk/packages/shared/src/feature-flags.ts +++ b/sdk/packages/shared/src/feature-flags.ts @@ -35,9 +35,8 @@ type Primitive = string | number | boolean | bigint | symbol | null | undefined; type HasNonPrimitiveFieldNames = { [K in keyof T]-?: Exclude extends never ? never : K; }[keyof T]; -type HasOnlyPrimitiveFields = HasNonPrimitiveFieldNames extends never - ? true - : false; +type HasOnlyPrimitiveFields = + HasNonPrimitiveFieldNames extends never ? true : false; export type FeatureFlagsContextPrimitiveValued = AssertTrue< HasOnlyPrimitiveFields >;