VisReason (Vision-Centric): leaderboard
Metric: Average accuracy (%) over the 10 reasoning categories (1,505 questions on everyday vision-centric reasoning: multiple-choice and fill-in-the-blank answers matched by regular expression, open-ended answers judged by GPT-5-mini, bounding boxes matched at IoU above 0.5 with partial credit), 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 June 2027. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) (Low) | 47.5 |
| 2 | GPT-5.2 (Thinking) | 35.5 |
| 3 | GPT-5 Mini (High) | 33.2 |
| 4 | GPT-5 (Low) | 33.2 |
| 5 | GPT-5 Mini (Medium) | 31.8 |
| 6 | GPT-5 Mini (Low) | 28.1 |
| 7 | O4 Mini | 27.3 |
| 8 | Qwen 3 VL 235B A22B (Thinking) | 26.7 |
| 9 | Qwen 3 VL 32B (Thinking) | 22.1 |
| 10 | GPT-5 Nano | 19 |
| 11 | Qwen 3 VL 8B (Thinking) | 19 |
| 12 | Qwen 3 VL 8B Instruct | 18.8 |
| 13 | Gemini 2.5 Flash | 16 |
| 14 | GPT-4o | 15.6 |
| 15 | Qwen 2.5 VL 32B Instruct | 15.4 |
Interactive version: theaggregate.ai/benchmark?slug=visreason-vision-centric · How It Works · Data refreshed daily, snapshot 2026-10-07.