AutoRLBench 38 Issue / 19 Solution - Issue: leaderboard
Metric: Root cause found, 38 failed training runs (%). Source: thinking-space.github.io. Saturation forecast: Around January 2028. 10 models tracked. Updated 18 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-6 Astra | Max | 31.6 | #7 |
| 2 | GPT-6.1 Sol | Max | 28.9 | #12 |
| 3 | GLM-5.3 | Max | 21.1 | #72 |
| 4 | Claude Fable 5.1 | Max | 15.8 | #17 |
| 5 | Kimi K3.1 Pro | Max | 15.8 | #32 |
| 6 | Grok 4.7 | Max | 15.8 | #48 |
| 7 | Gemini 4 Argon | Max | 15.8 | #14 |
| 8 | Qwen 3.8 Max | Max | 13.2 | #63 |
| 9 | Kimi K3 | Max | 0 | #60 |
| 10 | DeepSeek V4 Pro | Max | 0 | #194 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=autorlbench-38-issue-19-solution-issue · How It Works · Data refreshed daily, snapshot 2026-11-18.