FormationEval — leaderboard

Metric: Accuracy (%). Source: huggingface.co. 72 models tracked.

Top models

#ModelScore
1Gemini 3 Pro (Preview)99.8
2GLM-4.798.6
3Gemini 3 Flash (Preview)98.2
4Gemini 2.5 Pro97.8
5Grok 4.1 Fast97.6
6Kimi K2 (Thinking)97.2
7Claude Opus 4.597
8GPT-5 Mini (Medium)96.4
9DeepSeek R196.2
10Grok 4 Fast96
11GPT-5 Mini (High)95.6
12O4 Mini (High)95.2
13GPT-5 Mini (Low)95.2
14Gemini 2.5 Flash95
15Grok 3 Mini95

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