SpatialWorld - Digital Games: leaderboard
Metric: Task success rate (%) on the 105 digital 3D game tasks (Block3D, Maze, Maze3D, Rubik cube, Snake); task success rate (%) judged by terminal-state verifiers, vision-only egocentric observation and a unified text action interface, temperature 1.0, 30-turn history, step budget twice the human golden action count plus 10; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 39 |
| 2 | Gemini 3 Flash | 38.1 |
| 3 | GPT-5 | 36.4 |
| 4 | Gemini 2.5 Pro | 32.6 |
| 5 | Kimi K2.5 | 31 |
| 6 | Qwen 3 VL 235B A22B (Thinking) | 28.3 |
| 7 | Qwen 3.5 397B A17B | 26 |
| 8 | Seed 2.0 Lite | 24.8 |
| 9 | GLM-4.5V | 14.5 |
| 10 | GPT-5.4 | 11.9 |
| 11 | GLM-4.6V | 8.1 |
| 12 | Qwen 2.5 VL 72B Instruct | 7.6 |
| 13 | Qwen 3 VL 235B A22B Instruct | 5 |
Interactive version: theaggregate.ai/benchmark?slug=spatialworld-digital-games · How It Works · Data refreshed daily, snapshot 2026-09-29.