MCPEvol-Bench - Planning Effectiveness (Early Stage): leaderboard

Metric: Planning Effectiveness score (1 to 10 rubric; original servers; 201 synthesized multi-server tasks over a fixed set of candidate MCP servers from 123 real servers; rubric scores on a 1 to 10 scale from a DeepSeek-Chat judge over the instruction, final answer and execution trajectory (MCP-Bench evaluation framework)). Source: arxiv.org. Saturation forecast: Around December 2026. 12 models tracked.

Top models

#ModelScore
1Claude Sonnet 4.66.76
2Claude Opus 4.66.7
3Gemma 4 31B (IT)6.67
4Claude Sonnet 4.5 (Thinking)6.17
5GPT-5.45.71
6Qwen 3.5 27B5.57
7Qwen 3.5 9B4.89
8Gemini 2.5 Pro4.62
9GPT-5.14.35
10O33.79
11Llama 3.3 70B3.56
12GPT-4o2.85

Interactive version: theaggregate.ai/benchmark?slug=mcpevol-bench-planning-effectiveness-early-stage · How It Works · Data refreshed daily, snapshot 2026-09-29.