MLE-Bench Lite (OpenMLE-Evo): leaderboard
FrontisAI OpenRSI's MLE-Bench Lite run — Kaggle-competition ML engineering with the OpenMLE-Evo evolutionary program-search harness, 12-hour per-task budget on one RTX 4090; mean medal rate over the 22-task Lite split.
Metric: Medal Average (%). Source: frontisai.github.io. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5.2 | 62.12 | #80 |
| 2 | MiniMax-M3 | 59.09 | #151 |
| 3 | Qwen 3.6 35B A3B | 39.39 | #238 |
| 4 | Qwen 3 30B A3B 2507 (Thinking) | 34.85 | #379 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mle-bench-lite-openmle-evo · How It Works · Data refreshed daily, snapshot 2026-11-05.