SelfBench - getsentry/sentry: leaderboard
Coding agents implement changes rebuilt from the merged pull requests of the open-source repository getsentry/sentry: the pull request becomes the instruction and hidden tests written for it decide success. Each row is one model in one harness (Claude Code, Codex or pi) at one reasoning level. Task success rate, higher is better. Source: selfbench.dev.
Metric: Task success (%) on 78 tasks rebuilt from the repository's merged pull requests (hidden tests). Source: selfbench.dev. Saturation forecast: Around February 2027. 11 models tracked. Updated 13 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Claude Opus 5.5 | High | 82.05 | #7 |
| 2 | Claude Sonnet 5.5 | High | 82.05 | #21 |
| 3 | GPT-6.1 Sol | High | 80.77 | #12 |
| 4 | DeepSeek V4.1 Flash (1017) | High | 79.49 | #44 |
| 5 | GPT-6 Astra | Max | 78.21 | #8 |
| 6 | Kimi K3 | High | 76.92 | #59 |
| 7 | GLM-5.3 | High | 73.08 | #73 |
| 8 | Mistral Large 4 Preview | High | 69.23 | #206 |
| 9 | Claude Haiku 5.5 | High | 67.95 | #76 |
| 10 | DeepSeek V4.1 Flash | High | 57.69 | #69 |
| 11 | GPT-6 Luna | High | 51.28 | #105 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=selfbench-getsentry-sentry · How It Works · Data refreshed daily, snapshot 2026-11-17.