MMGist - Spatial Understanding: leaderboard

Metric: Accuracy (%) on the Spatial Understanding dimension (items kept from EmbSpatialBench and CountBench); 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 December 2026. 27 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview) (Medium)79.7
2Qwen 3.6 Plus79.3
3Qwen 3.6 35B A3B77.2
4Seed 2.0 Pro74.9
5Seed 2.0 Lite74.8
6Gemma 4 31B73.7
7Qwen 3.5 9B73.5
8Seed 2.0 Mini72
9Gemma 4 26B A4B70
10Qwen 3.5 4B68.7
11Gemini 3.1 Flash Lite (Medium)68.2
12Qwen 3.5 27B67.8
13Step3 VL 10B65.8
14GPT-5 (Medium)63.6
15GPT-5 Mini (Medium)60.5

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