PerfCodeBench: leaderboard
Metric: CGRE (self-reported). Source: benchmarklist.com. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 73.99 |
| 2 | Qwen 3.6 Max | 66.32 |
| 3 | GPT-5.4 | 66.27 |
| 4 | Claude Opus 4.5 | 65.09 |
| 5 | Qwen 3.6 27B | 65.07 |
| 6 | DeepSeek V4 Pro | 64.3 |
| 7 | Kimi K2.6 | 62.14 |
| 8 | Claude Sonnet 4.5 | 61.6 |
| 9 | Gemini 3.1 Pro (Preview) | 61.46 |
| 10 | DeepSeek V4 Flash | 61.25 |
| 11 | Seed 2.0 Lite | 60.19 |
| 12 | Qwen 3.6 Plus | 56.46 |
| 13 | Gemini 3.1 Flash | 45.89 |
| 14 | Kimi K2 | 44.73 |
| 15 | Qwen 3.6 35B A3B | 44.38 |
Interactive version: theaggregate.ai/benchmark?slug=perfcodebench · How It Works · Data refreshed daily, snapshot 2026-09-05.