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
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) (Medium) | 79.7 |
| 2 | Qwen 3.6 Plus | 79.3 |
| 3 | Qwen 3.6 35B A3B | 77.2 |
| 4 | Seed 2.0 Pro | 74.9 |
| 5 | Seed 2.0 Lite | 74.8 |
| 6 | Gemma 4 31B | 73.7 |
| 7 | Qwen 3.5 9B | 73.5 |
| 8 | Seed 2.0 Mini | 72 |
| 9 | Gemma 4 26B A4B | 70 |
| 10 | Qwen 3.5 4B | 68.7 |
| 11 | Gemini 3.1 Flash Lite (Medium) | 68.2 |
| 12 | Qwen 3.5 27B | 67.8 |
| 13 | Step3 VL 10B | 65.8 |
| 14 | GPT-5 (Medium) | 63.6 |
| 15 | GPT-5 Mini (Medium) | 60.5 |
Interactive version: theaggregate.ai/benchmark?slug=mmgist-spatial-understanding · How It Works · Data refreshed daily, snapshot 2026-09-29.