PerfCodeBench - Correct-and-Runnable Rate: leaderboard
Metric: Correct-and-runnable rate (%): share of tasks whose generated code compiles, runs and passes the oracle (compilation errors, runtime errors and wrong output all count as failures), 1,854 system-level performance-optimization tasks across C, C++, Go, Java, Python and CUDA; the model returns one replacement source file from the task metadata, interface contract and baseline in one shot; a candidate earns performance credit only if it compiles, runs and passes the task oracle; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 | 71.25 |
| 2 | Claude Opus 4.5 | 70.55 |
| 3 | Claude Sonnet 4.5 | 69.36 |
| 4 | GPT-5 | 61.81 |
| 5 | DeepSeek V4 Pro | 54.48 |
| 6 | Seed 2.0 Lite | 47.57 |
| 7 | Qwen 3.6 Max | 47.25 |
| 8 | DeepSeek V4 Flash | 47.09 |
| 9 | Qwen 3.6 Plus | 45.2 |
| 10 | Gemma 4 26B A4B | 45.15 |
| 11 | Gemma 4 31B | 42.45 |
| 12 | Kimi K2 | 36.03 |
| 13 | Qwen 3.6 35B A3B | 32.09 |
| 14 | Llama 4 Maverick | 29.4 |
| 15 | Gemini 3.1 Pro (Preview) | 25.57 |
Interactive version: theaggregate.ai/benchmark?slug=perfcodebench-correct-and-runnable-rate · How It Works · Data refreshed daily, snapshot 2026-10-07.