StreamPro-Bench - Object Understanding: leaderboard
The object understanding task of StreamPro-Bench (Li et al., 2026): the model tracks a named object and confirms its state changes as they happen, on a 1 FPS stream. StreamPro-F1 in percent, which credits a response only when it arrives inside the task time window and a Gemini 2.5 Pro judge accepts its content, with precision and recall penalizing extra and missed responses; higher is better.
Metric: StreamPro-F1 (%; a response counts only when a Gemini 2.5 Pro judge credits its content and it arrives inside the task time window, and precision and recall over the response trajectory penalize extra and missed responses; videos streamed at 1 FPS; confirm object state changes as they happen, within 3 seconds). Source: arxiv.org. Saturation forecast: Around July 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 VL 7B | 11.3 | #1796 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=streampro-bench-object-understanding · How It Works · Data refreshed daily, snapshot 2026-11-12.