MCP Atlas (agentic MCP tool-use benchmark) — Broad bucket (32 models). Not merged w/ MCPAtlas Public: non-overlap range (8.6-83.6 vs 66.6-73.8).
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Muse Spark 1.1 | Meta | 88.1 | 3 | 2026-08-24 |
| 2 | Opus 5 | Anthropic | 85.8 | 2 | 2026-08-24 |
| 3 | Claude Fable 5 | Anthropic | 84.7 | 3 | 2026-08-24 |
| 4 | Kimi K3 | Moonshot | 84.2 | 3 | 2026-08-24 |
| 5 | GPT-5.6 Sol | OpenAI | 83.6 | 3 | 2026-08-24 |
| 6 | Gemini 3.5 Flash | 83.6 | 5 | 2026-08-24 | |
| 7 | Claude Opus 4.8 | Anthropic | 83.6 | 8 | 2026-08-24 |
| 8 | GPT-5.5 | OpenAI | 82.8 | 7 | 2026-08-24 |
| 9 | GLM-5.2 Full Open Source | Z.ai | 82.6 | 6 | 2026-08-23 |
| 10 | Muse Spark | Meta | 82.2 | 1 | 2026-08-24 |
| 11 | Inkling-Small | Thinking Machines | 79.6 | 2 | 2026-08-24 |
| 12 | Claude Opus 4.7 | Anthropic | 79.1 | 7 | 2026-08-24 |
| 13 | Hy3-preview | Tencent | 79.1 | 1 | 2026-08-23 |
| 14 | Gemini 3.1 Pro | 78.2 | 13 | 2026-08-24 | |
| 15 | glm-5p2 | Z.ai | 77.8 | 1 | 2026-08-24 |
| 16 | Claude Opus 4.6 | Anthropic | 76.8 | 8 | 2026-08-24 |
| 17 | Qwen3.7 Max | Alibaba | 76.4 | 3 | 2026-08-23 |
| 18 | Kimi K2.7 Code | Moonshot | 76 | 2 | 2026-08-23 |
| 19 | Inkling | Thinking Machines | 76 | 3 | 2026-08-24 |
| 20 | GLM-5.1 | Z.ai | 75.6 | 6 | 2026-08-24 |
| 21 | muse-glimmer-30b | Meta | 75.5 | 1 | 2026-08-23 |
| 22 | MiniMax M3 | MiniMax | 74.2 | 3 | 2026-08-23 |
| 23 | DeepSeek-V4-Pro | DeepSeek | 74.2 | 5 | 2026-08-23 |
| 24 | Qwen3.6 Plus | Alibaba | 74.1 | 3 | 2026-08-23 |
| 25 | Qwen3.7 Plus Preview | Alibaba | 73.2 | 1 | 2026-08-23 |
| 26 | GPT-5.4 | OpenAI | 70.6 | 6 | 2026-08-24 |
| 27 | Gemini 3 Pro | 70.3 | 3 | 2026-08-24 | |
| 28 | Claude Opus 4.5 | Anthropic | 69.8 | 4 | 2026-08-24 |
| 29 | Claude Sonnet 4.6 | Anthropic | 69.5 | 4 | 2026-08-24 |
| 30 | GLM-5 | Z.ai | 69.2 | 4 | 2026-08-23 |
| 31 | DeepSeek-V4-Flash | DeepSeek | 69 | 1 | 2026-07-04 |
| 32 | Qwen3.5 27B | Alibaba | 68.4 | 1 | 2026-06-06 |
| 33 | Kimi K2.6 | Moonshot | 68.1 | 1 | 2026-07-16 |
| 34 | GPT-5.2 | OpenAI | 68 | 4 | 2026-08-24 |
| 35 | Kimi K2.5 | Moonshot | 64 | 4 | 2026-07-16 |
| 36 | Qwen3.6 35B A3B | Alibaba | 62.8 | 2 | 2026-08-23 |
| 37 | Qwen3.5 35B A3B | Alibaba | 62.4 | 1 | 2026-06-06 |
| 38 | DeepSeek-V3.2 | DeepSeek | 62.2 | 3 | 2026-05-18 |
| 39 | Gemini 3 Flash Preview | 62 | 3 | 2026-08-24 | |
| 40 | Claude Sonnet 4.5 | Anthropic | 59.5 | 2 | 2026-08-24 |
| 41 | GLM-4.7 | Z.ai | 58.1 | 2 | 2026-08-24 |
| 42 | GPT-5.4 mini | OpenAI | 57.7 | 2 | 2026-08-24 |
| 43 | Gemma 4 31B | 57.2 | 1 | 2026-06-06 | |
| 44 | Gemini 3.1 Flash Lite Preview | 57.1 | 1 | 2026-08-24 | |
| 45 | GPT-5.4 nano | OpenAI | 56.1 | 1 | 2026-08-23 |
| 46 | GPT-5.1 | OpenAI | 50.1 | 1 | 2026-08-24 |
| 47 | Gemma 4 26B A4B | 50 | 1 | 2026-06-06 | |
| 48 | MiniMax M2.7 | MiniMax | 48.8 | 2 | 2026-05-18 |
| 49 | o3-pro | OpenAI | 44.5 | 1 | 2026-08-24 |
| 50 | Claude Opus 4.1 | Anthropic | 40.9 | 1 | 2026-07-29 |
| 51 | Claude Haiku 4.5 | Anthropic | 40.2 | 1 | 2026-08-24 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.