GapEval - World Knowledge (Understanding): leaderboard

Metric: Accuracy (%; world-knowledge questions about entities shown in images; text answers to GapEval's bidirectional questions, judged correct or incorrect by GPT-5-mini against reference answers, mean of ten sampled runs). Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked.

Top models

#ModelScore
1Gemini 2.5 Flash97.96
2GPT-5 Mini97.28
3Qwen 3 VL 235B A22B96.6
4GPT-596.52
5Qwen 3 VL 8B94.3

Interactive version: theaggregate.ai/benchmark?slug=gapeval-world-knowledge-understanding · How It Works · Data refreshed daily, snapshot 2026-09-26.