MMOOC - Missing Knowledge and Background (Open-Ended VQA): leaderboard

Metric: Out-of-context score (%; mean of the refusal rate, the share of the Missing Knowledge and Background Open-Ended VQA questions, which the image and context cannot answer, that the model correctly declines, and refusal rationality, the mean 0-1 judge score of its abstention reasoning in steps of 0.25; judge scores averaged over GPT-5.6, Claude Opus 5 and DeepSeek-V4-Pro). Source: arxiv.org. Saturation forecast: Around 2030. 18 models tracked.

Top models

#ModelScore
1Gemma 4 26B80
2Gemma 4 31B70.5
3Llama 4 Maverick58
4Ministral 3 8B58
5GPT-4o57.75
6O153.75
7Claude Opus 4.646.5
8Qwen 3.5 122B A10B41.5
9O341.25
10Qwen 3.5 27B39
11Ministral 3 14B36.5
12InternVL3-8B32
13Gemini 3.1 Pro (Preview)20.25

Interactive version: theaggregate.ai/benchmark?slug=mmooc-missing-knowledge-and-background-open-ended-vqa · How It Works · Data refreshed daily, snapshot 2026-09-29.