# Scale Labs MCP Atlas (MCP Atlas)

> A Scale Labs public leaderboard mirrored as display-only reference data. It does not affect BenchLM rankings.

Canonical page: https://benchlm.ai/benchmarks/scale-mcp-atlas

- Category: [Agentic](/agentic)
- Last updated: September 29, 2026 snapshot

## About MCP Atlas

- Year: 2026
- Tasks: 34 published rows
- Format: Published Scale leaderboard score
- Difficulty: External agent and model evaluation
- Paper: [Scale Labs leaderboard](https://labs.scale.com/leaderboard/mcp_atlas)

BenchLM mirrors 34 published rows from the MCP Atlas public table captured on September 29, 2026 snapshot.

MCP Atlas is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (34 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [Muse Spark 1.1](/models/muse-spark-1-1) | Meta | 88.10% |
| 2 | [Fable 5.1](https://labs.scale.com/leaderboard/mcp_atlas) | Anthropic | 87.20% |
| 3 | [claude-opus-5 (xhigh)](https://labs.scale.com/leaderboard/mcp_atlas) | Anthropic | 85.80% |
| 4 | [Qwen3.8-2.4T-A95B (xHigh)](https://labs.scale.com/leaderboard/mcp_atlas) | Alibaba | 84.50% |
| 5 | [GLM-5.3](/models/glm-5-3) | Z.AI | 84.20% |
| 6 | [Gemini 3.5 Flash (high)](https://labs.scale.com/leaderboard/mcp_atlas) | Google | 83.60% |
| 7 | [Claude Fable 5](/models/claude-fable) | Anthropic | 83.30% |
| 8 | [kimi-k3 (max)](https://labs.scale.com/leaderboard/mcp_atlas) | Kimi | 82.30% |
| 9 | [claude-opus-4-8 (max)](https://labs.scale.com/leaderboard/mcp_atlas) | Anthropic | 82.20% |
| 10 | [Muse Spark](/models/muse-spark) | Meta | 82.20% |
| 11 | [gpt-5.6 (sol)](https://labs.scale.com/leaderboard/mcp_atlas) | OpenAI | 81.80% |
| 12 | [Inkling-Small](/models/inkling-small) | Thinking Machines Lab | 79.20% |
| 13 | [claude-opus-4-7 (max)](https://labs.scale.com/leaderboard/mcp_atlas) | Anthropic | 79.10% |
| 14 | [gemini-3.1-pro-preview (high)](https://labs.scale.com/leaderboard/mcp_atlas) | Google | 78.20% |
| 15 | [glm-5p2](https://labs.scale.com/leaderboard/mcp_atlas) | Z.AI | 77.80% |
| 16 | [claude-opus-4-6 (max)](https://labs.scale.com/leaderboard/mcp_atlas) | Anthropic | 76.80% |
| 17 | [Inkling (xHigh)](https://labs.scale.com/leaderboard/mcp_atlas) | Thinkingmachines | 76.00% |
| 18 | [glm-5p1](https://labs.scale.com/leaderboard/mcp_atlas) | Z.AI | 75.60% |
| 19 | [gpt-5.5 (xhigh)](https://labs.scale.com/leaderboard/mcp_atlas) | OpenAI | 75.30% |
| 20 | [gpt-5.4 (xhigh)](https://labs.scale.com/leaderboard/mcp_atlas) | OpenAI | 70.60% |
| 21 | [gemini-3-pro-preview](https://labs.scale.com/leaderboard/mcp_atlas) | Google | 70.30% |
| 22 | [claude-opus-4-5 (high)](https://labs.scale.com/leaderboard/mcp_atlas) | Anthropic | 69.80% |
| 23 | [Claude Sonnet 4.6](/models/claude-sonnet-4-6) | Anthropic | 69.50% |
| 24 | [gpt-5.2 (xhigh)](https://labs.scale.com/leaderboard/mcp_atlas) | OpenAI | 67.60% |
| 25 | [kimi-k2p5](https://labs.scale.com/leaderboard/mcp_atlas) | Kimi | 64.40% |
| 26 | [Nemotron 3 Ultra (thinking)](https://labs.scale.com/leaderboard/mcp_atlas) | Nvidia | 63.10% |
| 27 | [gemini-3-flash-preview](https://labs.scale.com/leaderboard/mcp_atlas) | Google | 62.00% |
| 28 | [claude-sonnet-4-5 (thinking)](https://labs.scale.com/leaderboard/mcp_atlas) | Anthropic | 59.50% |
| 29 | [glm-4p7](https://labs.scale.com/leaderboard/mcp_atlas) | Z.AI | 58.10% |
| 30 | [gemini-3.1-flash-lite (high)](https://labs.scale.com/leaderboard/mcp_atlas) | Google | 57.10% |
| 31 | [gpt-5.4-mini (xhigh)](https://labs.scale.com/leaderboard/mcp_atlas) | OpenAI | 56.70% |
| 32 | [gpt-5.1 (high)](https://labs.scale.com/leaderboard/mcp_atlas) | OpenAI | 50.10% |
| 33 | [o3-pro](/models/o3-pro) | OpenAI | 44.50% |
| 34 | [Claude Haiku 4.5](/models/claude-haiku-4-5) | Anthropic | 40.20% |

## FAQ

### What does MCP Atlas measure?

A Scale Labs public leaderboard mirrored as display-only reference data. It does not affect BenchLM rankings.

### Which model leads the published MCP Atlas snapshot?

Muse Spark 1.1 currently leads the published MCP Atlas snapshot with a score of 88.10%.

### How many models are evaluated on MCP Atlas?

The September 29, 2026 snapshot contains 34 AI models.

### Does MCP Atlas affect BenchLM's overall score?

Not directly. MCP Atlas is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.