KnowHal: leaderboard

Metric: Overall score (%; mean of the positive-question and negative-question average accuracies; 1,800 entity-image samples in 10 domains with paired positive and false-premise negative questions; positive answers scored by standard VQA answer normalization and soft matching, negative answers by a GPT-4o-mini judge (option matching for relation questions); zero-shot). Source: arxiv.org. Saturation forecast: Around 2030. 14 models tracked.

Top models

#ModelScore
1Gemini 3 Flash (Preview)67.9
2Qwen 3 VL 32B Instruct66.6
3Qwen 2.5 VL 72B Instruct66
4Seed 2.0 Mini61.1
5Kimi K2.560.5
6GPT-5.254.2
7GLM-4.6V54.1
8GPT-5.6 Sol52.1
9InternVL3.5-8B43.7

Interactive version: theaggregate.ai/benchmark?slug=knowhal · How It Works · Data refreshed daily, snapshot 2026-09-29.