STEMBench120 v1.0.0 (Closed Book) - Engineering: leaderboard

Metric: Mean fractional scientific progress (0–4; 20 tasks/domain; zero-credit failures included). Source: www.ulam.ai. Saturation forecast: Around January 2028. 4 models tracked.

Top models

#ModelScore
1Hy4 preview2.62
2Mistral Large 42.6
3DeepSeek V4.1 Flash2.54
4MiMo-V2.6-Pro2.17

Interactive version: theaggregate.ai/benchmark?slug=stembench120-v1-0-0-closed-book-engineering · How It Works · Data refreshed daily, snapshot 2026-10-10.