SlopCodeBench - Core Solved
Code · 2026-01-20
SlopCodeBench evaluates code at a sequence of development checkpoints. Core Solved is the percentage of checkpoints whose core tests pass.
Top models (higher is better)
| Model | Score |
|---|---|
| Opus 4.6 | 65.3 |
| GPT-5.5 | 65.3 |
| Opus 4.7 | 64.3 |
| GPT-5.4 | 61.2 |
| GPT-5.3-Codex | 59.2 |
| Opus 4.5 | 56.1 |
| Sonnet 4.6 | 56.1 |
| GPT-5.2-Codex | 54.6 |
| Composer 2 | 51.5 |
| GPT-5.4 Mini | 51.0 |
| Kimi K2.6 | 51.0 |
| Kimi K2.5 | 39.8 |
| GLM-5.1 | 38.8 |
| GPT-5.3-Codex-Spark | 29.1 |
| MiniMax M2.7 | 28.6 |