VisReason (Vision-Centric) - Open-Ended: leaderboard
VisReason Open-Ended scores multimodal LLMs on the 309 open-ended 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 309 open-ended questions (judged correct or not by GPT-5-mini against the reference answer) 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 February 2027. 18 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | Low | 72.5 | #119 |
| 2 | GPT-5 | Low | 61.5 | #178 |
| 3 | GPT-5 Mini | High | 60.2 | #364 |
| 4 | GPT-5.2 | Thinking | 52.1 | #185 |
| 5 | O4 Mini | 48.5 | #368 | |
| 6 | Qwen 3 VL 235B A22B (Thinking) | 43.7 | #459 | |
| 7 | GPT-5 Nano | 29.8 | #906 | |
| 8 | Qwen 3 VL 32B (Thinking) | 29.8 | #573 | |
| 9 | Qwen 3 VL 8B Instruct | 28.6 | #1029 | |
| 10 | Qwen 3 VL 30B A3B (Thinking) | 25.6 | #719 | |
| 11 | Qwen 3 VL 8B (Thinking) | 25.1 | #910 | |
| 12 | Qwen 2.5 VL 32B Instruct | 23.9 | #1062 | |
| 13 | Gemini 2.5 Flash | 12 | #463 | |
| 14 | InternVL3-14B | 11.7 | #1214 | |
| 15 | GPT-4o | 10.4 | #751 |
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-open-ended · How It Works · Data refreshed daily, snapshot 2026-11-21.