OVO-S-Bench (Streaming Mode): leaderboard
OVO-S-Bench (InternLM, 2026) for streaming video models in query-agnostic streaming mode: frames arrive one by one and memory decisions are made before the question is asked. The score is overall multiple-choice accuracy in percent, the mean of the four levels; higher is better.
Metric: Overall multiple-choice accuracy (%) on the 1,680-question evaluation set in query-agnostic streaming mode (frames fed sequentially at the model's published rate, memory decisions made before the question arrives), macro-average of the four levels. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | StreamForest-7B | 44.13 | |
| 2 | StreamingVLM-7B | 43.03 | |
| 3 | Flash-VStream-7B | 24.94 | #7335 |
Interactive version: theaggregate.ai/benchmark?slug=ovo-s-bench-streaming-mode · How It Works · Data refreshed daily, snapshot 2026-11-13.