deepseek-ai/DeepSeek-V4-Pro VRAM requirements

1599B params, Compressed recurrent state

DeepSeek-V4-Pro needs about 851 GB of GPU memory for FP8 weights. The smallest fitting setup at 32k context is 8× NVIDIA B200. It supports 32 concurrent conversations with SGLang defaults.

GPU requirements by context length

8k context
GPUGPUsConcurrent users
NVIDIA B200864
NVIDIA H200 SXM823
NVIDIA H100 SXM16 (verify server configuration)36
NVIDIA H100 NVL16 (verify server configuration)64
NVIDIA A100 80GB SXM16 (verify server configuration)24
NVIDIA RTX PRO 6000 Blackwell Server Edition16 (verify server configuration)64
NVIDIA H2016 (verify server configuration)64
NVIDIA L40Sno fit
32k context
GPUGPUsConcurrent users
NVIDIA B200832
NVIDIA H200 SXM86
NVIDIA H100 SXM16 (verify server configuration)9
NVIDIA H100 NVL16 (verify server configuration)17
NVIDIA A100 80GB SXM16 (verify server configuration)6
NVIDIA RTX PRO 6000 Blackwell Server Edition16 (verify server configuration)18
NVIDIA H2016 (verify server configuration)18
NVIDIA L40Sno fit
128k context
GPUGPUsConcurrent users
NVIDIA B20088
NVIDIA H200 SXM81
NVIDIA H100 SXM16 (verify server configuration)2
NVIDIA H100 NVL16 (verify server configuration)4
NVIDIA A100 80GB SXM16 (verify server configuration)1
NVIDIA RTX PRO 6000 Blackwell Server Edition16 (verify server configuration)4
NVIDIA H2016 (verify server configuration)4
NVIDIA L40Sno fit

FP8 weights, SGLang defaults, estimates.

Weights by precision

Weight memory
PrecisionWeightsSmallest fitting setup at 32k
FP8about 851 GB8x NVIDIA B200
INT4 (AWQ)about 851 GB8x NVIDIA B200

Model notes

Attention
Compressed recurrent state
Architecture
Mixture-of-Experts
Layers
61
Hidden size
7,168
Vocabulary
129,280
Native precision
bfloat16

This model keeps a fixed-size recurrent state per conversation instead of a growing token cache, so very long chats cost little extra memory.

Frequently asked questions

Will deepseek-ai/DeepSeek-V4-Pro run on a single H100?
No, a single H100 SXM cannot hold deepseek-ai/DeepSeek-V4-Pro at 32k context with FP8 weights. The model requires multiple GPUs.
What is the cheapest GPU setup for deepseek-ai/DeepSeek-V4-Pro?
The cheapest fitting setup at 32k context is 8x NVIDIA B200 with FP8 weights and SGLang defaults.
How much VRAM does deepseek-ai/DeepSeek-V4-Pro need at 128k context?
At 128k context on the cheapest fitting setup, deepseek-ai/DeepSeek-V4-Pro uses about 106 GB of VRAM per GPU for FP8 weights and about 6.5 GB per conversation for the KV cache. The total VRAM needed depends on the GPU count and parallelism configuration.
Can I run deepseek-ai/DeepSeek-V4-Pro with INT4 quantization?
Yes, deepseek-ai/DeepSeek-V4-Pro has INT4 weight figures. Its weights take about 851 GB in INT4 versus about 851 GB in FP8. INT4 can make the model fit on fewer GPUs, but check inference quality for your workload.

Want a different context length, GPU, or quantization? The calculator runs the same numbers live, preloaded with this model.

Open in calculator

Don't want to run this yourself? We deploy and operate it for you. Book a call →

Related models

google/gemma-4-31B-it zai-org/GLM-4.7 zai-org/GLM-5.1