# MCPMark

Tests AI agents on tasks that require using MCP tool servers such as Notion and GitHub.

Category: agents. Type: leaderboard. Status: active.
Metric: Pass@1: share of tasks solved on first try (%) (higher is better).
Source: MCPMark team — https://mcpmark.ai/leaderboard/verified, results as of 2026-07-20, retrieved 2026-11-17, license: not stated.

| Rank | Model | Pass@1 (%) |
|---|---|---|
| 1 | kimi-k3-max | 96.06 |
| 2 | gpt-5-5-xhigh | 92.91 |
| 3 | gpt-5-6-sol-max | 92.91 |
| 4 | claude-fable-5-max | 86.61 |
| 5 | kimi-k2-7-code | 81.89 |
| 6 | claude-opus-4-8-max | 76.38 |
| 7 | kimi-k2-6 | 72.83 |
| 8 | deepseek-v4-pro-max | 71.65 |

Notes: Verified leaderboard, 8 models; ranks 2 and 3 tie on score but are listed in page order.
