SelfBench - supabase/supabase: leaderboard
Metric: Task success (%) on 36 tasks rebuilt from the repository's merged pull requests (hidden tests). Source: selfbench.dev. Saturation forecast: Around July 2027. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-6.1 Sol (High) | 77.78 |
| 2 | Claude Opus 5.5 (High) | 77.78 |
| 3 | Kimi K3 (High) | 77.78 |
| 4 | Claude Sonnet 5.5 (High) | 77.78 |
| 5 | GPT-6.1 Sol (Low) | 75 |
| 6 | Claude Opus 5.5 (Low) | 75 |
| 7 | GPT-6 Luna (High) | 66.67 |
| 8 | GPT-6 Luna (Low) | 44.44 |
Interactive version: theaggregate.ai/benchmark?slug=selfbench-supabase-supabase · How It Works · Data refreshed daily, snapshot 2026-10-03.