SlopCodeBench: leaderboard
SlopCodeBench (scbench.ai) — resisting 'slop' patterns in generated code: isolated-solved rate on tasks demanding clean, minimal solutions.
Metric: Isolated Solved (%). Source: www.scbench.ai. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT 5.5 (High) | 28.06 | |
| 2 | GPT 5.3-Codex (High) | 26.02 | |
| 3 | GPT 5.4 (High) | 25.51 | |
| 4 | GPT 5.2-Codex (High) | 21.94 | |
| 5 | Opus 4.6 (High) | 20.92 | |
| 6 | Opus 4.7 (High) | 20.92 | |
| 7 | Kimi K2.6 (High) | 18.88 | #104 (Kimi K2.6) |
| 8 | Opus 4.5 (High) | 17.35 | |
| 9 | Sonnet 4.6 (High) | 16.84 | |
| 10 | Composer 2 (None) | 16.33 |
Interactive version: theaggregate.ai/benchmark?slug=slopcodebench · How It Works · Data refreshed daily, snapshot 2026-10-20.