E-Bench-Code: leaderboard

Metric: Avg@3 (%; per-trial success averaged over tasks and 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)81.11
2Kimi K3 (Max)77.61
3GPT-5.5 (xHigh)77.19
4Grok 4.5 (High)69.24
5Hy364.4
6Gemini 3.5 Flash (High)62.54
7Qwen 3.7 Max61.92
8GLM-5.2 (Max)60.99
9Seed 2.1 Pro53.04
10DeepSeek V4 Pro (Max)47.68
11MiniMax-M346.85

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