~ / models / open-weights / videojev
VideoJev
Nanjing University's System One model for video understanding and interaction: temporal grounding and video QA as typed decisions over a shared video state, with the paper reporting >100× QA speedup over autoregressive video LLMs at parity. The video-native entrant the class was missing. Apache 2.0, ~309 stars.
- section
- Models
- group
- Open weights
- source
- github.com
- repo
- mcg-nju/videojev
- stars
- 310 ▕███████░░░░░▏
- 7-day gain
- +34
- forks
- 12
- language
- Python
- license
- Apache-2.0
- last pushed
- oct 18
- added to list
- nov 12
## history
2026-10-112 → 310 stars2026-11-14
2026-11-1312 → 12 forks2026-11-14
2026-11-130 → 0 open issues2026-11-14