AutoRLBench 45 Issue / 19 Solution - Solution: leaderboard
Solution-implementation split (19 tasks) of Thinking-Space AutoRLBench 45-issue / 19-solution edition measuring whether agents implement working fixes that pass held-out RL training checks; pass rate (%).
Metric: Repaired run passes its rerun, 19 tasks (%). Source: thinking-space.github.io. Saturation forecast: Around 2029. 10 models tracked. Updated 21 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-6.1 Sol | Max | 31.6 | #15 |
| 2 | Qwen 3.8 Max | Max | 31.6 | #63 |
| 3 | GPT-6 Astra | Max | 26.3 | #10 |
| 4 | DeepSeek V4.1 Flash | Max | 26.3 | #67 |
| 5 | GLM-5.3 | Max | 26.3 | #72 |
| 6 | Claude Fable 5.1 | Max | 26.3 | #20 |
| 7 | Kimi K3.1 Pro | Max | 26.3 | #36 |
| 8 | Grok 4.7 | Max | 26.3 | #51 |
| 9 | Kimi K3 | Max | 15.8 | #61 |
| 10 | DeepSeek V4 Pro | Max | 5.3 | #196 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=autorlbench-45-issue-19-solution-solution · How It Works · Data refreshed daily, snapshot 2026-11-21.