MCP-Bench: leaderboard
Benchmark for evaluating LLM agents on complex real-world tool-use tasks through MCP servers, covering schema understanding, LLM-judged task completion, tool usage, and planning effectiveness.
Metric: Overall Score (%). Source: github.com. Source: huggingface.co. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 74.9 | #98 |
| 2 | O3 | 71.5 | #124 |
| 3 | GPT-OSS-120B | 69.2 | #341 |
| 4 | Gemini 2.5 Pro | 69 | #152 |
| 5 | Claude Sonnet 4 | 68.1 | #199 |
| 6 | Qwen 3 235B A22B 2507 | 67.8 | #260 |
| 7 | GLM-4.5 | 66.8 | #275 |
| 8 | GPT-OSS-20B | 65.4 | #518 |
| 9 | Kimi K2 | 62.9 | #243 |
| 10 | Qwen 3 30B A3B 2507 Instruct | 62.7 | #488 |
| 11 | Gemini 2.5 Flash Lite | 59.8 | #422 |
| 12 | GPT-4o | 59.5 | #343 |
| 13 | Gemma 3 27B (IT) | 58.2 | #523 |
| 14 | Llama 3.3 70B Instruct | 55.8 | #534 |
| 15 | GPT-4o Mini | 55.7 | #599 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mcp-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.