CodegenBench - LeetSunway: leaderboard
Metric: Pass@1 (%): share of problems whose first generated program compiles, runs and matches the reference output within tolerance; zero-shot through the official APIs with a short optimization system prompt, on the 20 LeetSunway kernels that must use the master-slave core paradigm of the SW26010 processor (Sunway TaihuLight, swgcc); higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.7 | 48 |
| 2 | Claude Opus 4.6 | 33 |
| 3 | Qwen 3.6 Plus | 22 |
| 4 | Claude Sonnet 4.6 | 6 |
| 5 | DeepSeek V4 Flash | 5 |
| 6 | DeepSeek V4 Pro | 0 |
| 7 | DeepSeek V3.2 | 0 |
| 8 | Qwen 3.5 Plus | 0 |
| 9 | Qwen 3.6 Flash | 0 |
Interactive version: theaggregate.ai/benchmark?slug=codegenbench-leetsunway · How It Works · Data refreshed daily, snapshot 2026-09-29.