WeirdML: leaderboard
17 unusual ML tasks requiring genuine understanding - not memorization. Models generate PyTorch code in Docker containers with 5 iterative debug rounds. Tests shape recognition, chess prediction, and more.
Metric: Average Score. Source: htihle.github.io. Status: saturation imminent. 158 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5.1 (Max) | 92.9 |
| 2 | Claude Fable 5.1 (High) | 92.3 |
| 3 | Claude Fable 5 (Max) | 91.94 |
| 4 | Claude Opus 5 (Max) | 91.78 |
| 5 | Claude Opus 5 (High) | 91.59 |
| 6 | GPT-5.6 Sol (High) | 88.76 |
| 7 | Claude Fable 5 (High) | 87.85 |
| 8 | GPT-5.6 Sol (Max) | 86.97 |
| 9 | GPT-5.5 (xHigh) | 84.91 |
| 10 | GPT-5.5 (High) | 83.9 |
| 11 | Claude Opus 4.8 (xHigh) | 82.89 |
| 12 | Kimi K3 (Max) | 82.57 |
| 13 | GPT-5.6 Terra (High) | 78.27 |
| 14 | Claude Opus 4.6 (High) | 77.95 |
| 15 | GPT-5.3 Codex (xHigh) | 77.9 |
Interactive version: theaggregate.ai/benchmark?slug=weirdml · How It Works · Data refreshed daily, snapshot 2026-09-05.