deepseek-ai
DeepSeek V4 Pro (0813)
Frontier 1.6T-parameter (49B active) MoE reasoning model with 1M-token context. The 2026-08-13 official release that supersedes the V4-Pro preview.MITHugging Face
No server in stock right now has the 960 GB this model needs for inference.
GPU memory it needs
From model cards and vendor documentation. Lower precision needs less memory, with slightly lower quality. Estimates are marked.
| What you want to do | Precision | Minimum | Recommended | Setup | Basis |
|---|---|---|---|---|---|
| Run the model | FP8, 8-bit | 960 GB | 1128 GB | 8x H200 (vLLM recipe default, single-node) | Published |