OVO-S-Bench - Generative Spatial Reasoning: leaderboard
Metric: Level 3 (generative spatial reasoning: simulation, consistency checks, route planning) multiple-choice accuracy (%) on the 1,680-question evaluation set, prefix-access mode (128 frames sampled uniformly from the video before the query timestamp, question given with the frames), mean of its three task families. Source: arxiv.org. Saturation forecast: Around 2032. 31 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 235B A22B | 61.2 |
| 2 | Qwen 3.5 397B A17B | 58.07 |
| 3 | Gemini 3.1 Pro (Preview) | 55.9 |
| 4 | Qwen 3 VL 4B Instruct | 54.54 |
| 5 | Gemini 3.1 Flash Lite | 54.14 |
| 6 | Qwen 3.5 27B (Non-reasoning) | 52.39 |
| 7 | Qwen 3 VL 32B Instruct | 51.94 |
| 8 | GPT-5.4 | 50.8 |
| 9 | Qwen 3.5 9B (Non-reasoning) | 50.2 |
| 10 | Qwen 3.5 4B (Non-reasoning) | 49.25 |
| 11 | Grok 4.1 Fast (Non-reasoning) | 48.49 |
| 12 | InternVL3.5-8B | 47.2 |
| 13 | Qwen 2.5 VL 7B Instruct | 45.86 |
| 14 | Gemma 4 26B A4B | 45.01 |
| 15 | Gemma 4 E4B | 42.8 |
Interactive version: theaggregate.ai/benchmark?slug=ovo-s-bench-generative-spatial-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-29.