DeepSeek V4-Pro — VRAM, Speed & Local Setup

Written by Jakub Rusinowski · Last updated April 24, 2026

Model library → DeepSeek V4 → DeepSeek V4-Pro

DeepSeek's flagship MoE — 1.6T total, 49B active, 1M context, MIT license. Not realistically self-hostable on any consumer or prosumer setup: even Q2 quantization is estimated around 400GB, requiring an 8x80GB+ H100/H800/B200-class server. Included for reference; the API or DeepSeek's hosted chat is the practical access path for almost everyone.

DeepSeek V4-Pro needs about 967 GB of VRAM at Q2 (experimental, datacenter only) — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.

Specifications

Parameters	1.6 Trillion (49B active)
Context window	1,000,000
Architecture	Mixture-of-Experts
Provider	DeepSeek
Licence	MIT
Specified at	Q2 (experimental, datacenter only)
System RAM	1024 GB
Record updated	2026-04-24

Licence

MIT — commercial use permitted. Commercial use permitted. No usage restrictions beyond attribution.

VRAM and Speed by Quantization

Modelled on a reference NVIDIA RTX 4090 (24 GB), with no KV cache (this record has no published architecture). Speed figures are ESTIMATES from the memory-bandwidth roofline described on the methodology page, not benchmarks we ran — rows marked measured come from published or reader-submitted runs. VRAM here includes the KV cache, so it reads higher than the headline figure above, which does not.

Quant	Weights	VRAM needed	Est. speed	Fit on 24 GB
Q2_K	526.0 GB	526.8 GB	—	Won't fit
Q3_K_M	682.0 GB	682.8 GB	—	Won't fit
Q4_K_M	966.0 GB	966.8 GB	—	Won't fit
Q5_K_M	1134.0 GB	1134.8 GB	—	Won't fit
Q6_K	1312.0 GB	1312.8 GB	—	Won't fit
Q8_0	1700.0 GB	1700.8 GB	—	Won't fit
F16	3200.0 GB	3200.8 GB	—	Won't fit

Want the memory numbers alone, at every quantization level and your own context length? Use the DeepSeek V4-Pro VRAM calculator.

Deploy in the Cloud NowRunPod

or compare on Vast.ai

As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.

How to Run DeepSeek V4-Pro

Install Ollama, then run:

ollama run deepseek-v4

Weights on Hugging Face: deepseek-ai/DeepSeek-V4-Pro.

Best for: frontier tasks, enterprise, research, cloud api.

Can I Run DeepSeek V4-Pro on My GPU?

Other DeepSeek V4 Sizes

DeepSeek V4-Flash — about 172 GB VRAM at Q4_K_M

DeepSeek V4-Pro — Frequently Asked Questions

How much VRAM does DeepSeek V4-Pro need?

About 967 GB at Q2 (experimental, datacenter only) — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.

Does DeepSeek V4-Pro run on an RTX 4090 (24 GB)?

No. DeepSeek V4-Pro needs about 967 GB at Q2 (experimental, datacenter only), more than a single RTX 4090's 24 GB. It needs a larger card, several GPUs, or Apple Silicon with enough unified memory — or it runs with part of the weights offloaded to system RAM, which is much slower.

How do I run DeepSeek V4-Pro locally?

Install Ollama and run `ollama run deepseek-v4`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.

What other sizes does DeepSeek V4 come in?

DeepSeek V4-Flash (172 GB), DeepSeek V4-Pro (967 GB). Every size shares the family's training and licence; the larger ones score higher and need proportionally more memory.

← All DeepSeek V4 models | VRAM calculator | Check your own hardware