VisReason (Vision-Centric) - Inductive Reasoning: leaderboard
Metric: Accuracy (%) on the Inductive Reasoning category (conceptual reasoning: infer underlying rules from observed instances and apply them to novel cases) of VisReason, answers scored by format (regular-expression match, GPT-5-mini judge for open-ended answers, IoU above 0.5 for boxes), 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 December 2027. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) (Low) | 53.1 |
| 2 | Qwen 3 VL 235B A22B (Thinking) | 46.2 |
| 3 | GPT-5.2 (Thinking) | 42.3 |
| 4 | GPT-5 Mini (High) | 41.8 |
| 5 | GPT-5 Mini (Medium) | 41.8 |
| 6 | GPT-5 (Low) | 40.7 |
| 7 | Qwen 3 VL 32B (Thinking) | 38.6 |
| 8 | GPT-5 Mini (Low) | 37.5 |
| 9 | O4 Mini | 36.4 |
| 10 | Qwen 3 VL 8B (Thinking) | 35 |
| 11 | Qwen 3 VL 30B A3B (Thinking) | 34.9 |
| 12 | Qwen 3 VL 8B Instruct | 33.7 |
| 13 | Gemini 2.5 Flash | 31.6 |
| 14 | GPT-5 Nano | 29.8 |
| 15 | Qwen 2.5 VL 32B Instruct | 25.1 |
Interactive version: theaggregate.ai/benchmark?slug=visreason-vision-centric-inductive-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-07.