PostTrainBench — leaderboard

AI R&D automation benchmark: agents get 4 small LLMs, one H100, and 10 hours to improve model performance through post-training. Measures dataset curation, training strategy, and constraint awareness.

Metric: Weighted Avg Score. Source: posttrainbench.com. Status: saturation imminent. 12 models tracked.

Top models

#ModelScore
1Claude Opus 4.623.2
2Gemini 3.1 Pro (Preview)21.6
3GPT-5.221.4
4GPT-5.420.2
5GPT-5.1 Codex Max19.7
6Gemini 3 Pro18.1
7GPT-5.3 Codex17.8
8GPT-5.2 Codex17.2
9Claude Opus 4.517.1
10Claude Sonnet 4.616.4
11GLM-513.9
12Claude Sonnet 4.59.9

Interactive version: theaggregate.ai/benchmark?slug=posttrainbench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.