DeepSeek

DeepSeek V4 Pro

Name: DeepSeek V4 Pro
Author: DeepSeek

Frontier

HuggingFace

5.9MDownloads4.5KLikesApr 2026Veröffentlicht1.0M TokenKontextMITLizenz100 HerausragendQualität

DeepSeek V4 Pro (1600B parameters) requires approximately 865.4 GB of VRAM with NVFP4 quantization. As a Mixture of Experts model with 49B active parameters, it uses less memory than its total parameter count suggests. For the best balance of quality and speed, we recommend hardware with at least 996 GB of VRAM.

Loslegen

— kopieren & einfügen, um lokal auszuführen

Copy-paste commands to run DeepSeek V4 Pro on your machine.

Run

docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
  --hf-repo "deepseek-ai/DeepSeek-V4-Pro" \
  --hf-file "DeepSeek-V4-Pro-NVFP4.gguf" \
  -c 4096 -ngl 99

Quick specs

Parameters1600B (49B active)

Architecturemoe (MoE)

Context1.0M tokens

Modalitytext

Min RAM624 GB

Rec. RAM896 GB (NVFP4)

LicenseMIT

FamilyDeepSeek

✓ Code✓ Reasoning

About this model

DeepSeek V4 Pro is a 1.6T-parameter sparse MoE (49B active, 384 routed + 1 shared expert) built for million-token agentic reasoning. Experts ship natively in FP4, so the real on-disk footprint is roughly 862 GB (FP4 experts + FP8 attention) rather than the trillion-scale FP16 size — but it is still a server/workstation deployment: realistic local use targets 8x 80GB GPUs or 1 TB+ unified memory, and at long Think Max contexts the KV cache dominates.

•1.6T total / 49B active sparse MoE — 384 routed + 1 shared expert
•Native FP4 experts: ~862 GB on disk, not trillion-scale FP16
•1M-token context for million-token agent workflows
•Server/workstation class — use distills or the Flash variant for local

Verwandte Modelle

Quantisierungsoptionen

VRAM-Schätzungen nach Quantisierungsstufe

No hardware detected — fit column shows raw VRAM estimates

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	624.0 GB	Low	—
Q3_K_S	3	784.0 GB	Low	—
NVFP4	4	896.0 GB	Medium	—
Q4_K_M	4	976.0 GB	Medium	—
Q5_K_M	5	1152.0 GB	High	—
Q6_K	6	1312.0 GB	High	—
Q8_0	8	1712.0 GB	Very High	—
F16	16	3280.0 GB	Maximum	—

Quality benchmarks

DeepSeek V4 Pro benchmark scores

Benchmark verified

Coding

SWE-bench Verified80.6%

HumanEval+—

Aider Polyglot—

LiveCodeBench93.5%

Reasoning

MMLU-Pro87.5%

GPQA Diamond—

MATH-500—

ARC Challenge—

Source: vendor-reported · 2026-04-24

Hardware-Kompatibilität

Eignungsschätzungen für alle Hardware

Rechner öffnen

Computing compatibility...

Speicheraufschlüsselung

Reference: RTX 2060 6GB

Weights862.0 GB

KV Cache1.9 GB

Runtime0.9 GB

Headroom0.6 GB

Häufig gestellte Fragen

FAQ — DeepSeek V4 Pro

Siehe auch

Quantisierungsleitfaden Bewertungsmethodik Rechner öffnen