SmellBench - Partial Component: leaderboard

SmellBench (2026) scores coding-agent CLIs on repairing 65 expert-validated architectural code smells in scikit-learn. This board averages the per-task score over the 13 smells the expert judged partially valid, where repairs earn credit by how much they reduce the smell; range -1 to 1, higher is better.

Metric: Mean per-task score s (-1 to 1: 1 for a full repair or a correctly flagged false positive, partial credit for a severity reduction, negative for a worsened smell or a genuine smell flagged as false positive) over the 13 smells the expert judged partially valid, 65 expert-validated hard-severity PyExamine architectural smells in scikit-learn, off-the-shelf coding agent CLIs with the SmellBench agent skill and GEPA-optimized task packets, default reasoning settings, one run per agent; higher is better. Source: arxiv.org. Saturation forecast: Around February 2028. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.3 Codex Spark0.353
2GPT-5.40.158#138
3GPT-5.3 Codex0.147#125
4Claude Opus 4.60.09#114
5Gemini 3.1 Pro (Preview)0.056#102
6Claude Sonnet 4.60.056#179
7Claude Haiku 4.50.042#587
8GPT-5.1 Codex Max0.037#212
9Mistral Vibe + Devstral 20.007
10Gemini 3.1 Flash Lite (Preview)0#419

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=smellbench-partial-component · How It Works · Data refreshed daily, snapshot 2026-11-22.