LongCat-Flash-Omni
560B-total (27B active) open omni-modal model with 128K context for real-time audio-visual interaction and streaming speech.
Extends LongCat-Flash with lightweight modality encoders and a speech decoder, trained with curriculum-style progressive multimodal training. MIT.
- Date
- Thursday, 23 October 2025
- Lab
- Meituan (LongCat)
- Kind
- open-weights
- Access
- open weights
Figures
| Measure | Value | Measured by |
|---|---|---|
| Total / active parameters | 560B / 27B | company |
| Context length | 128K | company |
Date from HF initial commit (2025-10-23 UTC); Wikipedia confirms the model exists.
Sources
This record was checked against its sources on 6 October 2026. How we check