SWE-PRBench (Config A) - Detection Rate: leaderboard
Metric: Detection rate (%): share of the human-flagged issues that the model's comments catch (judge-confirmed and matched one to one), on a stratified 100-PR sample of SWE-PRBench (40 direct, 40 contextual and 20 latent-difficulty pull requests from active open-source repositories), each model asked at temperature 0 for 4 to 6 line-anchored review comments as JSON, a GPT-5.2 judge matching them to the human reviewers' comments (confirmed, plausible or fabricated), with the diff-only context (task focus, key-change summary, diff and minimal metadata; about 2,000 tokens); higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V3 | 31.2 | #312 |
| 2 | Claude Haiku 4.5 | 30.6 | #271 |
| 3 | Mistral Large 3 | 30.5 | #388 |
| 4 | Claude Sonnet 4.6 | 29.7 | #85 |
| 5 | Mistral Small | 25.7 | #700 |
| 6 | Llama 3.3 70B | 22.3 | #569 |
| 7 | GPT-4o | 22 | #333 |
| 8 | GPT-4o Mini | 21 | #588 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=swe-prbench-config-a-detection-rate · How It Works · Data refreshed daily, snapshot 2026-10-11.