MCP-Universe (Agent Track) — leaderboard

MCP-Universe agent track benchmarks autonomous agents (not raw LLMs) on complex enterprise workflows including 3D design and financial analysis.

Metric: Overall Success Rate. Source: mcp-universe.github.io. 8 models tracked.

Top models

#ModelScore
1ReAct (GPT-5-Medium)43.72
2OpenAI Agent SDK (o3-Medium)31.6
3HarmonyReAct (GPT-OSS-120B)31.17
4ReAct (Claude-4.0-Sonnet)29.44
5Cursor Agent 1.4.5 (Claude-4.0-Sonnet-Thinking)26.41
6ReAct (o3-Medium)26.41
7HarmonyReAct (GPT-OSS-20B)24.24
8OpenAI Agent SDK (GPT-OSS-120B)11.26

Interactive version: theaggregate.ai/benchmark?slug=mcp-universe-agent-track · How the rankings work · Data refreshed daily, snapshot 2026-07-22.