MLS-Bench (Vanilla) - Language Models: leaderboard

MLS-Bench (Vanilla) Language Models scores the first proposal each model tests on the 18 Language Models tasks of MLS-Bench (2026), where a model must invent a better ML method inside a frozen research codebase. Task scores are normalized so the worst reproduced human baseline is 0 and the best is 50, on a 0 to 100 scale where higher is better.

Metric: Mean normalized task score (0-100): each metric is anchored so the worst reproduced human baseline scores 0 and the best 50 (a metric's theoretical optimum maps to 100), metrics are averaged with human weights within a setting and by geometric mean across at least three generalization settings, over the 18 MLS-Bench tasks of the Language Models area; Vanilla protocol: the score of the model's first tested proposal (tools: edit, test, submit, undo); high reasoning effort with a 10,000-token thinking budget, web search disabled, fixed seeds; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 5 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3.1 Pro (Preview)High39#99
2DeepSeek V3.2Thinking35.3#419
3Claude Opus 4.6Thinking26.9#113
4GPT-5.4High24.1#135
5Qwen 3.6 PlusThinking22#238

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=mls-bench-vanilla-language-models · How It Works · Data refreshed daily, snapshot 2026-11-21.