MMOOC - Image-Question Mismatch (Open-Ended VQA): leaderboard

Metric: Shifted in-context score (%; mean of accuracy on the answerable Image-Question Mismatch Open-Ended VQA questions and answer rationality, the mean 0-1 judge score of the answer's correctness and evidence consistency 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 December 2026. 18 models tracked.

Top models

#ModelScore
1Qwen 3.5 27B89
2Qwen 3.5 122B A10B86.75
3O385.25
4O181
5Gemma 4 31B80.5
6Ministral 3 8B80.5
7GPT-4o80.25
8Llama 4 Maverick80
9Gemma 4 26B78
10Ministral 3 14B73.5
11InternVL3-8B70.75
12Gemini 3.1 Pro (Preview)52
13Claude Opus 4.635.75

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