MMOOC - Unclear Logical and Symbolic (Open-Ended VQA): leaderboard

Metric: Out-of-context score (%; mean of the refusal rate, the share of the Unclear Logical and Symbolic 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 2032. 18 models tracked.

Top models

#ModelScore
1Gemma 4 26B58
2GPT-4o54
3Gemma 4 31B54
4O147.5
5Qwen 3.5 122B A10B42
6Qwen 3.5 27B40.25
7Ministral 3 8B37.25
8Ministral 3 14B36.5
9O334.5
10Llama 4 Maverick31.75
11Claude Opus 4.623.5
12InternVL3-8B15.5
13Gemini 3.1 Pro (Preview)9.5

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