OVO-S-Bench - Context Tracking: leaderboard
Level 2 of OVO-S-Bench (InternLM, 2026), a streaming spatial intelligence benchmark: multiple-choice accuracy in percent on tracking spatial evidence that has left the view, with 128 frames sampled from the video before the query time. Higher is better.
Metric: Level 2 (spatiotemporal context tracking of evidence no longer in view) 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 October 2028. 31 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 64.04 | #94 |
| 2 | GPT-5.4 | 57.59 | #123 |
| 3 | Qwen 3.5 397B A17B | 55.39 | #246 |
| 4 | Qwen 3 VL 235B A22B | 55.18 | #479 |
| 5 | Qwen 3.5 27B (Non-reasoning) | 55.15 | #359 (Qwen 3.5 27B) |
| 6 | Gemini 3.1 Flash Lite | 52.19 | #364 |
| 7 | Qwen 3 VL 32B Instruct | 51.91 | #585 |
| 8 | Qwen 3.5 9B (Non-reasoning) | 49.37 | #803 (Qwen 3.5 9B) |
| 9 | Qwen 3.5 4B (Non-reasoning) | 48.2 | #1175 (Qwen 3.5 4B) |
| 10 | Qwen 3 VL 4B Instruct | 48.18 | #1279 |
| 11 | Grok 4.1 Fast (Non-reasoning) | 46.59 | #430 (Grok 4.1 Fast) |
| 12 | Gemma 4 26B A4B | 46.55 | #615 |
| 13 | Qwen 2.5 VL 7B Instruct | 45.51 | #1833 |
| 14 | Gemma 4 E4B | 42.76 | #2214 |
| 15 | Gemma 4 E2B | 36.52 | #3892 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=ovo-s-bench-context-tracking · How It Works · Data refreshed daily, snapshot 2026-11-13.