DeepSeek · open-weights model
DeepSeek-V4-Flash
DeepSeek-V4-Flash is a 284B-parameter model (MoE, 13B active) by DeepSeek. Its smallest quant (UD-Q2_K_XL) needs at least 98.0GB of GPU-addressable memory at 8K context; it loads cleanly at a quality quant on 2 of the 50 machines we track.
- Parameters
- 284B total · 13B active
- Layers
- 43
- Max context
- 1024K
- Attention
- HYBRID
- License
- mit
Config from deepseek-ai/DeepSeek-V4-Flash
Quants and what they need
Measured file sizes; “needs” = weights + KV cache at 8K context + overhead.
| Quant | File size | Needs ≥ |
|---|---|---|
| UD-Q8_K_XL | 161.9 GB | 163.1 GB |
| UD-Q4_K_XL | 155.1 GB | 156.3 GB |
| UD-Q3_K_XL | 129.4 GB | 130.7 GB |
| UD-Q2_K_XL | 96.8 GB | 98.0 GB |
What runs DeepSeek-V4-Flash
Every machine we track, graded at its best clean-fit quant.
Verdicts computed from published specs and measured GGUF sizes — how the math works.