MCPEvol-Bench - Task Fulfillment (Early Stage): leaderboard

Metric: Task Fulfillment 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
1GPT-5.47.23
2Claude Sonnet 4.67.22
3Claude Opus 4.67.15
4Claude Sonnet 4.5 (Thinking)6.4
5GPT-5.16.28
6O35.79
7Gemini 2.5 Pro5.73
8GPT-4o5.24
9Gemma 4 31B (IT)5.05
10Qwen 3.5 27B4.46
11Llama 3.3 70B3.84
12Qwen 3.5 9B3.38

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