MLS-Bench Lite: leaderboard
Official 30-task subset of MLS-Bench for evaluating whether AI systems can invent generalizable and scalable machine-learning methods.
Metric: Score. Source: mls-bench.com. Status: years away from saturation. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5 | 49.9 |
| 2 | Kimi K3 | 48.3 |
| 3 | GPT-5.6 Sol | 46.2 |
| 4 | Claude Opus 4.8 | 42.8 |
| 5 | Qwen 3.8 Max | 41 |
| 6 | GLM-5.2 | 40.4 |
| 7 | GPT-5.5 | 35.5 |
| 8 | Kimi K2.7 Code | 35.1 |
| 9 | Qwen 3.7 Max | 31.7 |
| 10 | Kimi K2.6 | 26.7 |
Interactive version: theaggregate.ai/benchmark?slug=mls-bench-lite · How It Works · Data refreshed daily, snapshot 2026-09-05.