MCP Atlas leaderboard
- Muse Spark 1.1 leads
- 88.1%
- models measured
- 32
- lead over #2
- +0.9 pts
- median
- 75.8%
Leader on MCP Atlas · read October 22
On October 22, 2026, 32 of our catalog's models had a MCP Atlas result, and Meta Muse Spark 1.1 led with 88.1%, ahead of Anthropic Claude Fable 5.1 at 87.2%.
MCP Atlas tests how well a model handles real-world tool use through the Model Context Protocol. The number is the share of tasks solved, as Scale AI publishes it.
Role in the index
MCP Atlas is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | Muse Spark 1.1Meta | 88.1% | 70.3 | $4.25 | |
| 2 | Claude Fable 5.1Anthropic | 87.2% | 84.2 | $50.00 | |
| 3 | Claude Opus 5Anthropic | 85.8% | 83.0 | $25.00 | |
| 4 | Qwen3.8 2.4T A95BQwen | 84.5% | Not benchmarked | $6.00 | |
| 5 | GLM 5.3Z.ai | 84.2% | 70.6 | $4.40 | |
| 6 | Gemini 3.5 FlashGoogle | 83.6% | 39.8 | $9.00 | |
| 7 | Claude Fable 5Anthropic | 83.3% | 79.3 | $50.00 | |
| 8 | Kimi K3Moonshot AI | 82.3% | 62.3 | $15.00 | |
| 9 | Claude Opus 4.8Anthropic | 82.2% | 65.0 | $25.00 | |
| 10 | GPT-5.6 SolOpenAI | 81.8% | 73.5 | $20.00 | |
| 11 | T Inkling SmallThinking Machines | 79.2% | 27.3 | $1.20 | |
| 12 | Claude Opus 4.7Anthropic | 79.1% | 67.5 | $25.00 | |
| 13 | Gemini 3.1 Pro PreviewGoogle | 78.2% | 32.3 | $12.00 | |
| 14 | GLM 5.2Z.ai | 77.8% | 61.1 | $4.40 | |
| 15 | Claude Opus 4.6Anthropic | 76.8% | 57.2 | $25.00 | |
| 16 | T InklingThinking Machines | 76.0% | 31.5 | $4.05 | |
| 17 | GLM 5.1Z.ai | 75.6% | 49.2 | $4.40 | |
| 18 | GPT-5.5OpenAI | 75.3% | 71.3 | $30.00 | |
| 19 | GPT-5.4OpenAI | 70.6% | 64.4 | $15.00 | |
| 20 | Claude Opus 4.5Anthropic | 69.8% | 44.0 | $25.00 | |
| 21 | Claude Sonnet 4.6Anthropic | 69.5% | 57.3 | $15.00 | |
| 22 | GPT-5.2OpenAI | 67.6% | 55.9 | $14.00 | |
| 23 | Kimi K2.5Moonshot AI | 64.4% | 38.6 | $2.85 | |
| 24 | Nemotron 3 UltraNVIDIA | 63.1% | 30.0 | $2.40 | |
| 25 | Gemini 3 Flash PreviewGoogle | 62.0% | 43.7 | $3.00 | |
| 26 | Claude Sonnet 4.5Anthropic | 59.5% | 39.5 | $15.00 | |
| 27 | GLM 4.7Z.ai | 58.1% | 55.1 | $2.20 | |
| 28 | Gemini 3.1 Flash LiteGoogle | 57.1% | Not benchmarked | $1.50 | |
| 29 | GPT-5.4 MiniOpenAI | 56.7% | 35.8 | $4.50 | |
| 30 | GPT-5.1OpenAI | 50.1% | 40.4 | $10.00 | |
| 31 | o3 ProOpenAI | 44.5% | Not benchmarked | $80.00 | |
| 32 | Claude Haiku 4.5Anthropic | 40.2% | 35.6 | $5.00 |
Only models in our catalog are listed, and each appears once with its best published run. 66 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out.