MCP Atlas
Scale AI's large-scale tool-use benchmark: 1,000 expert-written natural-language tasks over 36 real Model Context Protocol (MCP) servers and 220+ tools, requiring agents to discover and orchestrate multi-step tool calls; scored by pass rate via an LLM judge.
What is MCP Atlas?
Scale AI's large-scale tool-use benchmark: 1,000 expert-written natural-language tasks over 36 real Model Context Protocol (MCP) servers and 220+ tools, requiring agents to discover and orchestrate multi-step tool calls; scored by pass rate via an LLM judge. evals.report tracks reported MCP Atlas scores with the model, source, status, date, and run caveats attached — official leaderboard scores, vendor-reported launches, and clearly labeled community runs.
Top reported MCP Atlas score: Muse Spark 1.1 — 88.1% (pass rate).
| Model | Lab | Score↓ | Source model | Status | Date | |
|---|---|---|---|---|---|---|
| Muse Spark 1.1 | Meta | 88.1% | Muse Spark 1.1 | Verified | Jul 9, 2026 | Details |
| GLM-5.2Open | Z.ai | 76.8% | GLM-5.2 | Verified | Jun 16, 2026 | Details |
| MiniMax M3Open | MiniMax | 74.2% | MiniMax M3 | Verified | Jun 1, 2026 | Details |
| InklingOpen | Thinking Machines Lab | 74.1% | Inkling | Verified | Jul 15, 2026 | Details |
| DeepSeek V4 FlashOpen | DeepSeek | 69.0% | DeepSeek-V4-Flash (Think-Max) | Verified | Apr 24, 2026 | Details |
| Claude Opus 4.5 | Anthropic | 62.3% | Claude Opus 4.5 | Official | Nov 24, 2025 | Details |
| Gemini 3 Pro | Google DeepMind | 54.1% | Gemini 3 Pro | Official | Nov 18, 2025 | Details |
| GPT-5 | OpenAI | 44.5% | GPT-5 | Official | Aug 7, 2025 | Details |
| Claude Sonnet 4.5 | Anthropic | 43.8% | Claude Sonnet 4.5 | Official | Sep 29, 2025 | Details |
| Claude Opus 4.1 | Anthropic | 40.9% | Claude Opus 4.1 | Official | Aug 5, 2025 | Details |
| Claude Sonnet 4 | Anthropic | 35.6% | Claude Sonnet 4 | Official | May 22, 2025 | Details |
| Amazon Nova 2 Lite | Amazon | 24.6% | Nova 2 Lite | Verified | Dec 2, 2025 | Details |
| Kimi K2 InstructOpen | Moonshot AI | 23.9% | Kimi K2 Instruct | Official | Jul 11, 2025 | Details |
| Qwen3 235B A22B Instruct 2507Open | Alibaba / Qwen | 12.0% | Qwen3-235B-A22B | Official | Jul 21, 2025 | Details |
| Gemini 2.5 Pro | Google DeepMind | 8.8% | Gemini 2.5 Pro | Official | Mar 25, 2025 | Details |
| GPT-4o | OpenAI | 7.2% | GPT-4o | Official | May 13, 2024 | Details |
| Gemini 2.5 Flash | Google DeepMind | 3.4% | Gemini 2.5 Flash | Official | Apr 17, 2025 | Details |
Each row reports the model’s pass rate on MCP Atlas. Click a row for the full run context.