MMGist - Visual Perception: leaderboard

Metric: Accuracy (%) on the Visual Perception dimension (items kept from BLINK, HallusionBench and RealWorldQA); MMGist: 7,262 curated items from 18 vision-language benchmarks after removing items answerable without the image, items nearly every model solves and items with faulty labels; eight samples per item at temperature 1.0, step-by-step answer in a boxed block, 16,384 output tokens, medium effort where configurable; higher is better. Source: arxiv.org. Saturation forecast: Around May 2028. 27 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview) (Medium)75.4
2Seed 2.0 Pro70.2
3Qwen 3.6 Plus67.3
4Seed 2.0 Lite65.4
5Seed 2.0 Mini63.2
6Gemini 3.1 Flash Lite (Medium)62.4
7Qwen 3.6 35B A3B61.7
8GPT-5 (Medium)58.5
9Qwen 3.5 27B58
10Qwen 3.5 9B57.6
11Gemma 4 31B55.8
12Qwen 3.5 4B55.6
13Step3 VL 10B53.3
14GPT-5 Mini (Medium)52.7
15Gemma 4 26B A4B52.1

Interactive version: theaggregate.ai/benchmark?slug=mmgist-visual-perception · How It Works · Data refreshed daily, snapshot 2026-09-29.