E-Bench-Code - Pass^3: leaderboard

Metric: Pass^3 (%; tasks solved in all three trials; domain MCP tools plus an exec_code tool) on 323 synthetic state-changing tasks in three product environments (Honor of Kings, QQ Music, Tencent Meeting), graded by exact database-state diff; three trials per task, each model at its highest available thinking effort in a shared MCP harness. Source: arxiv.org. Saturation forecast: Around December 2026. 11 models tracked.

Top models

#ModelScore
1Claude Opus 4.8 (Max)68.66
2GPT-5.5 (xHigh)66.9
3Kimi K3 (Max)65.8
4Grok 4.5 (High)55.75
5Qwen 3.7 Max44.01
6Hy344.01
7GLM-5.2 (Max)42.25
8Gemini 3.5 Flash (High)40.49
9Seed 2.1 Pro30.99
10DeepSeek V4 Pro (Max)25.35
11MiniMax-M321.13

Interactive version: theaggregate.ai/benchmark?slug=e-bench-code-pass-pow-3 · How It Works · Data refreshed daily, snapshot 2026-09-29.