DeepSeek-V4-Flash
284B MoE · 13B active
~154 tok/s per stream
DetailsHide details
Where the memory goes
- Model weights143.59 GB
- KV cache0.24 GB
- Activations0.03 GB
- CUDA & framework2.49 GB
- Precision
- INT4 / AWQ
- Serving preset
- 8K context × 4 users
- Users at this context
- 752
- Spare context in leftover VRAM
- 3.6M
- Fit score
- 66/100 · High