MLE-bench: leaderboard
OpenAI machine-learning engineering benchmark where agents solve Kaggle-style ML tasks and are scored by medal rate across competition environments.
Metric: Medal Rate (%). Source: github.com. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Famou-Agent 2.0 | 64.44 | |
| 2 | AIBuildAI | 63.11 | |
| 3 | CAIRMARS+ | 62.67 | |
| 4 | MLEvolve | 61.33 | |
| 5 | PiEvolve(Fractal AI Research) | 61.33 | |
| 6 | Famou-Agent 2.0 | 59.56 | |
| 7 | ML-Master 2.0 | 56.44 | |
| 8 | CAIRMARS | 56 | |
| 9 | PiEvolve(Fractal AI Research) | 52 | |
| 10 | Leeroo | 50.67 |
Interactive version: theaggregate.ai/benchmark?slug=mle-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.