CommitSuite - Commit Classification: leaderboard

Ten-way commit type classification on CommitSuite (2026), a benchmark of Conventional Commits compliant commits from 243 open-source repositories in seven languages: the model reads a commit and assigns one of ten types (feat, fix, perf, style, refactor, docs, test, ci, build, chore) on a balanced 1,160-commit evaluation set. Accuracy in percent; higher is better.

Metric: Accuracy (%, times 100) of ten-way Conventional Commits type classification (feat, fix, perf, style, refactor, docs, test, ci, build, chore) on the balanced 1,160-commit Ten-category-eval set of CommitSuite (116 human-checked commits per type), zero-shot prompting with the commit diff and context; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1DeepSeek R163.8#482
2Gemini 2.0 Flash62.9#723
3Claude 3.7 Sonnet60.9#470
4GPT-4.160.8#505
5DeepSeek V359.7#729
6GPT-4o Mini55.9#1667

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=commitsuite-commit-classification · How It Works · Data refreshed daily, snapshot 2026-11-21.