SlopCodeBench - Erosion Mean
Code · 2026-01-20
Mean code-quality erosion, scaled to percent; lower is better.
Top models (lower is better)
| Model | Score |
|---|---|
| GPT-5.5 | 49.4 |
| MiniMax M2.7 | 50.0 |
| GPT-5.4 | 50.8 |
| GLM-5.1 | 55.0 |
| GPT-5.3-Codex-Spark | 58.6 |
| GPT-5.3-Codex | 64.4 |
| GPT-5.4 Mini | 65.5 |
| Opus 4.5 | 69.1 |
| Kimi K2.5 | 69.2 |
| Composer 2 | 71.6 |
| GPT-5.2-Codex | 72.8 |
| Opus 4.6 | 73.7 |
| Sonnet 4.6 | 74.1 |
| Opus 4.7 | 75.9 |
| Kimi K2.6 | 76.4 |