SelfBench - pingdotgg/t3code: leaderboard
Metric: Task success (%) on 39 tasks rebuilt from the repository's merged pull requests (hidden tests). Source: selfbench.dev. Saturation forecast: Around October 2027. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5.5 (High) | 76.92 |
| 2 | GPT-6.1 Sol (Low) | 66.67 |
| 3 | GPT-6.1 Sol (High) | 64.1 |
| 4 | Claude Sonnet 5.5 (High) | 58.97 |
| 5 | Kimi K3 (High) | 56.41 |
| 6 | GLM-5.3 (High) | 51.28 |
| 7 | GPT-6 Luna (High) | 46.15 |
| 8 | GPT-6 Luna (Low) | 25.64 |
Interactive version: theaggregate.ai/benchmark?slug=selfbench-pingdotgg-t3code · How It Works · Data refreshed daily, snapshot 2026-10-03.