What GPU do you need to run DeepSeek V4 Flash?
284B params 1M max context Tool useReasoning
The smallest GPU memory that runs DeepSeek V4 Flash well is 192 GB (Q4_K_M, e.g. Ryzen AI Max 400 (Gorgon Halo)). 4 of 132 GPUs run it well.
MMLU 86.4 SWE-bench Verified 79
VRAM needed per quantization
| Quantization | Weights | 16K context | 32K context | Full 1M |
|---|---|---|---|---|
| Q4_K_M | 161.0 GB | 162.3 GB | 163.6 GB | 245.0 GB |
| Q8_0 | 282.0 GB | 283.3 GB | 284.6 GB | 366.0 GB |
Weights plus fp16 KV cache. Leave ~1 GB headroom for the inference engine.
GPUs that can run DeepSeek V4 Flash
Best quantization that fits on each GPU, with the context and speed it gets. For chips with several memory sizes, the smallest that works is shown.
AMD
- Ryzen AI Max 400 (Gorgon Halo) 192 GB Q4_K_M 378K context · ~2 tok/s