benchmarks.site

Public leaderboard · Agents, tool use & computer use

MCPMark

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

Leaderboard

Top 3 right now

  1. 1stkimi-k3-max96.06
  2. 2ndgpt-5-5-xhigh92.91
  3. 3rdgpt-5-6-sol-max92.91

What it measures

Realistic multi-step tasks carried out through MCP servers for common services, verified programmatically.

How to read the numbers

Pass@1: share of tasks solved on first try (%) — higher is better.

Results

8 rows · higher is better (Pass@1 (%))CSV
1kimi-k3-max96.06
2gpt-5-5-xhigh92.91
3gpt-5-6-sol-max92.91
4claude-fable-5-max86.61
5kimi-k2-7-code81.89
6claude-opus-4-8-max76.38
7kimi-k2-672.83
8deepseek-v4-pro-max71.65

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

Source: MCPMark team · mcpmark.ai · results as of 2026-07-20 · retrieved 2026-11-17

Markdown · More Agents, tool use & computer use