SenseBench - Single-Image Perception: leaderboard
Metric: Exact-Match Accuracy (%; mean of the whether, what and how question types, single images). Source: arxiv.org. Saturation forecast: Around April 2027. 29 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 75.75 |
| 2 | Qwen 3.5 Plus | 63.29 |
| 3 | GPT-5.4 | 62.48 |
| 4 | Qwen 3 VL 8B | 56.98 |
Interactive version: theaggregate.ai/benchmark?slug=sensebench-single-image-perception · How It Works · Data refreshed daily, snapshot 2026-09-25.