VisReason (Vision-Centric) - Multiple-Choice: leaderboard

VisReason Multiple-Choice scores multimodal LLMs on the 615 multiple-choice questions of VisReason (2026), a benchmark of vision-centric reasoning in everyday images, pooled over its 10 reasoning categories. The score is accuracy in percent, higher is better.

Metric: Accuracy (%) on the 615 multiple-choice questions (option index extracted by regular expression) of VisReason, pooled over the 10 reasoning categories, zero-shot with chain-of-thought prompt templates, one per answer format; rows not shaded gray in the paper use explicit reasoning (GPT-5 and Gemini 3 Pro at low reasoning effort); higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 18 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3 Pro (Preview)Low47.8#119
2Qwen 3 VL 235B A22B (Thinking)40.8#459
3GPT-5.2Thinking39.4#185
4GPT-5 MiniMedium36.4#364
5Qwen 3 VL 32B (Thinking)34.5#573
6GPT-5Low34#178
7Gemini 2.5 Flash31.5#463
8O4 Mini30.9#368
9Qwen 3 VL 30B A3B (Thinking)30.9#719
10Qwen 3 VL 8B (Thinking)30.6#910
11Qwen 3 VL 8B Instruct27.8#1029
12GPT-4o25.2#751
13GPT-5 Nano23.6#906
14Qwen 2.5 VL 32B Instruct22.9#1062
15InternVL3-8B22.1#1782

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=visreason-vision-centric-multiple-choice · How It Works · Data refreshed daily, snapshot 2026-11-21.