DeepSeek/Mixture of Experts

DeepSeek V4 Pro 0813

Name: DeepSeek V4 Pro 0813
Author: DeepSeek

DeepSeek V4 Pro 0813 — the official release of V4 Pro, replacing the April preview: same 1.6T / 49B-active architecture with much stronger agentic and coding results. The checkpoint weighs 1.65T because it ships a DSpark speculative-decoding draft module on top of the base model.

chatcodingreasoningmultilingualmathagentictool_use

1650.5B

Parameters (49B active)

1024K

Context length

Benchmarks

Quantizations

152K

HF downloads

Architecture

MoE

Released

2026-08-13

Layers

KV Heads

Head Dim

512

Family

deepseek

Quantization Options

Context length:

Quant	Bits	VRAM @ 16K	Quality
IQ2_XXS	2.38	491.6 GB 491.5 + 0.1 KV	low
IQ2_M	2.93	605.1 GB 605.0 + 0.1 KV	low
Q2_K	3.16	652.5 GB 652.4 + 0.1 KV	low
IQ3_XXS	3.25	671.1 GB 671.0 + 0.1 KV	low
IQ3_XS	3.5	722.7 GB 722.6 + 0.1 KV	low
Q3_K_S	3.64	751.6 GB 751.5 + 0.1 KV	low
IQ3_M	3.76	776.3 GB 776.2 + 0.1 KV	low
Q3_K_M	4	825.9 GB 825.7 + 0.1 KV	low
Q3_K_L	4.3	887.7 GB 887.6 + 0.1 KV	moderate
IQ4_XS	4.46	920.8 GB 920.6 + 0.1 KV	moderate
Q4_K_S	4.67	964.1 GB 964.0 + 0.1 KV	moderate
Q4_K_M	4.89	1009.5 GB 1009.4 + 0.1 KV	good
Q5_K_S	5.57	1149.8 GB 1149.6 + 0.1 KV	good
Q5_K_M	5.7	1176.6 GB 1176.5 + 0.1 KV	good
Q6_K	6.56	1354.0 GB 1353.9 + 0.1 KV	excellent
Q8_0	8.5	1754.3 GB 1754.1 + 0.1 KV	lossless
FP16	16	3301.6 GB 3301.5 + 0.1 KV	lossless

Select your GPU above to see speed estimates and compatibility for each quantization.

Too big for a single GPU — plan a multi-GPU deployment

Even the lightest quant needs ~492 GB. Size GPUs, replicas, TCO and scaling for a production setup. Open in Enterprise →

▸ READY TO RUN THIS?RENT BY THE HOUR

RENT A GPU AND RUN DEEPSEEK V4 PRO 0813 NOW

Rent on Vast.ai →Or RunPod →

Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.

Community Ratings

Loading ratings...

Benchmarks (6)

GPQA Diamond92.8

AA Long Context80.3

Terminal-Bench 2.178.7

SciCode51.0

HLE41.0

Terminal-Bench 414.1

Run this model

▸Easiest way to get started·Beginners

DOCS ↗

curl -fsSL https://ollama.com/install.sh | sh

$ollama run deepseek:1651b-q4_K_M

Tag may need adjustment — check ollama.com/library/deepseek for available tags.

▸ SETUP GUIDE

Auto-setup with fitmyllm CLI

Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.

pip install fitmyllmthen run fitmyllmLearn more

Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines

HuggingFace GGUF Downloads Build Hardware

Find the best GPU for DeepSeek V4 Pro 0813

Build Hardware for DeepSeek V4 Pro 0813

QUANT	BPW	VRAM	QUALITY
IQ2_XXS	2.38	491.5 GB	65%
IQ2_M	2.93	605.0 GB	75%
Q2_K	3.16	652.4 GB	78%
IQ3_XXS	3.25	671.0 GB	82%
IQ3_XS	3.5	722.6 GB	84%
Q3_K_S	3.64	751.5 GB	85%
IQ3_M	3.76	776.2 GB	86%
Q3_K_M	4	825.7 GB	88%
Q3_K_L	4.3	887.6 GB	90%
IQ4_XS	4.46	920.6 GB	92%
Q4_K_S	4.67	964.0 GB	93%
Q4_K_M	4.89	1009.4 GB	94%
Q5_K_S	5.57	1149.6 GB	96%
Q5_K_M	5.7	1176.5 GB	96%
Q6_K	6.56	1353.9 GB	97%
Q8_0	8.5	1754.1 GB	100%
FP16	16	3301.5 GB	100%