Hardware guide · October 2026
What GPU do you need to run DeepSeek V4 Flash?
DeepSeek V4 Flash is a 284B (13B active) mixture-of-experts model. Here is exactly how much memory it needs and which GPUs and Macs run it well.
The short answer
To run DeepSeek V4 Flash at FP4 with 8K tokens of context you need about 168GB of VRAM (162.0GB of weights plus KV cache and overhead). No single consumer GPU can hold it. The most affordable setup is 8× RTX 3090 (~25 tokens/sec). On a Mac or other unified-memory system, the M3 Ultra 512GB runs it at ~32 tokens/sec.
- VRAM at FP4
- ~168GB
- GPUs that run it
- 3
- Parameters
- 284B
- CPU only (est.)
- ~5.9 tok/s
Our recommendations
The best GPUs for DeepSeek V4 Flash
Multi-GPU
8× RTX 3090192GB total · ~$8,400 used
- Quant
- FP4
- Speed (est.)
- ~25 tok/s
Split across cards with llama.cpp, vLLM, or ExLlama. Needs a board and PSU that can feed every card.
Faster multi-GPU
8× RTX 4090192GB total · ~$20,000 used
- Quant
- FP4
- Speed (est.)
- ~27 tok/s
Split across cards with llama.cpp, vLLM, or ExLlama. Needs a board and PSU that can feed every card.
Mac / unified memory
M3 Ultra 512GB512GB unified · $9,499 MSRP
- Quant
- FP4
- Speed (est.)
- ~32 tok/s
Lots of memory per dollar and silent, but slower than a GPU with the same capacity.
We may earn a commission when you buy through links on this page. It never changes what we recommend. How we make money
Memory
DeepSeek V4 Flash VRAM requirements
DeepSeek V4 Flash is released in FP4, so its memory use depends mostly on context length.
| Quantization | Quality | Weights | Total @ 8K ctx | Total @ 32K ctx |
|---|---|---|---|---|
| FP4 | NativeDefault | 162 GB | 168 GB | 168 GB |
Context
How context length changes the memory you need
Every token in the context window is stored in the KV cache. At FP4, with a 16-bit KV cache, DeepSeek V4 Flash needs:
| Context | KV cache | Total memory | With 8-bit KV |
|---|---|---|---|
| 4K tokens | 0.0 GB | 167.5 GB | 167.5 GB |
| 8K tokens | 0.0 GB | 167.5 GB | 167.5 GB |
| 16K tokens | 0.1 GB | 167.5 GB | 167.5 GB |
| 32K tokens | 0.2 GB | 167.6 GB | 167.5 GB |
| 64K tokens | 0.3 GB | 167.8 GB | 167.6 GB |
| 128K tokens | 0.7 GB | 168.1 GB | 167.8 GB |
| 256K tokens | 1.3 GB | 168.8 GB | 168.1 GB |
Compressed attention: layers alternate between compressed sparse attention (about 4 tokens merged per cache entry) and heavily compressed attention (about 128 tokens per entry), so the cache grows far more slowly than standard or MLA attention. DeepSeek's published figures put a full 1M-token cache at roughly 5GB, about 5,000 bytes per token. Treat this as approximate; actual use depends on runtime support.
Compatibility
Every GPU that runs DeepSeek V4 Flash
Each GPU at the highest-quality quantization that fits with 8K context, fastest first. Speeds are single-user estimates.
| GPU | Memory | Quant | Speed | Fit | Price |
|---|---|---|---|---|---|
| M5 Ultra 512GB | 512GB unified | FP4 | ~45 tok/s | Runs well | — |
| M5 Ultra 256GB | 256GB unified | FP4 | ~45 tok/s | Runs well | ~$9,499 |
| M3 Ultra 512GB | 512GB unified | FP4 | ~32 tok/s | Runs well | $9,499 |
Tips
Getting the most out of DeepSeek V4 Flash
- Budget memory for context. The KV cache grows with every token. Quantizing it to 8-bit roughly halves that cost with little quality loss.
- Offload experts, not layers. For mixture-of-experts models, llama.cpp's
--n-cpu-moe(or-ottensor overrides) keeps attention on the GPU and moves expert weights to system RAM. That keeps speed usable when the model does not fit in VRAM. - Pick the right runtime. Ollama and LM Studio are easiest. llama.cpp gives the most control. vLLM or SGLang are best when several people or agents share one machine.
Want to try different quantizations, context lengths, or GPU counts? Use our VRAM calculator, or read the DeepSeek V4 Flash overview for setup commands and benchmarks.
Frequently asked questions
How much VRAM do I need to run DeepSeek V4 Flash?
About 168GB at FP4 with an 8K context window. DeepSeek V4 Flash ships in FP4, so there are no smaller official quantizations. Longer contexts add 0.01GB per 1K tokens with a 16-bit KV cache.
Can I run DeepSeek V4 Flash on a 24GB GPU?
No. DeepSeek V4 Flash needs about 168GB at FP4, which does not fit on a 24GB card like the RTX 3090 or RTX 4090, even with CPU offload on a 64GB system.
Can I run DeepSeek V4 Flash on a Mac?
Yes. With 512GB of unified memory or more (for example the M3 Ultra 512GB), macOS can give the GPU enough memory to run it at FP4, at about ~32 tokens/sec. MLX builds are usually fastest on Apple silicon.
Can I run DeepSeek V4 Flash on CPU only, without a GPU?
Technically yes on a server with enough RAM, but at about 168GB it needs 128GB+ of RAM and will be slow on a desktop.