FormationEval: leaderboard

FormationEval — recognising football/soccer team formations and tactical structure from match data; accuracy.

Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 72 models tracked.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro (Preview)99.8#66
2GLM-4.798.6#192
3Gemini 3 Flash (Preview)98.2#82
4Gemini 2.5 Pro97.8#152
5Grok 4.1 Fast97.6#217
6Kimi K2 (Thinking)97.2#243 (Kimi K2)
7Claude Opus 4.597#83
8GPT-5 Mini (Medium)96.4#186 (GPT-5 Mini)
9DeepSeek R196.2#251
10Grok 4 Fast96#254
11GPT-5 Mini (High)95.6#186 (GPT-5 Mini)
12O4 Mini (High)95.2#180 (O4 Mini)
13GPT-5 Mini (Low)95.2#186 (GPT-5 Mini)
14Gemini 2.5 Flash95#245
15Grok 3 Mini95#281

No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.

Interactive version: theaggregate.ai/benchmark?slug=formationeval · How It Works · Data refreshed daily, snapshot 2026-10-20.