FitMyLLM
DeepSeek/Mixture of Experts

DeepSeekDeepSeek-V3 684.5B

chat
684.5B
Parameters (37B active)
160K
Context length
2
Benchmarks
17
Quantizations
928K
HF downloads
Architecture
MoE
Released
2025-01-20
Layers
61
KV Heads
128
Head Dim
56
Family
deepseek

Quantization Options

QuantBitsVRAMQuality
IQ2_XXS2.38204.1 GBlow
IQ2_M2.93251.2 GBlow
Q2_K3.16270.9 GBlow
IQ3_XXS3.25278.6 GBlow
IQ3_XS3.5300.0 GBlow
Q3_K_S3.64311.9 GBlow
IQ3_M3.76322.2 GBlow
Q3_K_M4342.7 GBlow
Q3_K_L4.3368.4 GBmoderate
IQ4_XS4.46382.1 GBmoderate
Q4_K_S4.67400.1 GBmoderate
Q4_K_M4.89418.9 GBgood
Q5_K_S5.57477.1 GBgood
Q5_K_M5.7488.2 GBgood
Q6_K6.56561.8 GBexcellent
Q8_08.5727.8 GBlossless
FP16161369.5 GBlossless

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

Benchmarks (2)

Arena Elo1373
BigCodeBench50.0

Run this model

Easiest way to get starteddocs →
curl -fsSL https://ollama.com/install.sh | sh
$ollama run deepseek-v3:684.5b-instruct-q4_k_m

Downloads and runs automatically. Add --verbose for speed stats.

Setup guide
HuggingFaceOllama LibraryGGUF DownloadsBuild Hardware

Find the best GPU for DeepSeek-V3 684.5B

Build Hardware for DeepSeek-V3 684.5B

DeepSeek-V3 684.5B — 684.5B Parameter Mixture of Experts LLM

Model Specifications

Parameters
684.5B (37B active)
Architecture
Mixture of Experts
Context Length
160K tokens
Capabilities
chat
Release Date
2025-01-20
Provider
DeepSeek
Family
deepseek

VRAM Requirements

QuantizationBPWVRAMQuality
IQ2_XXS2.38204.1 GB65%
IQ2_M2.93251.2 GB75%
Q2_K3.16270.9 GB78%
IQ3_XXS3.25278.6 GB82%
IQ3_XS3.5300.0 GB84%
Q3_K_S3.64311.9 GB85%
IQ3_M3.76322.2 GB86%
Q3_K_M4342.7 GB88%
Q3_K_L4.3368.4 GB90%
IQ4_XS4.46382.1 GB92%
Q4_K_S4.67400.1 GB93%
Q4_K_M4.89418.9 GB94%
Q5_K_S5.57477.1 GB96%
Q5_K_M5.7488.2 GB96%
Q6_K6.56561.8 GB97%
Q8_08.5727.8 GB100%
FP16161369.5 GB100%

Benchmark Scores

BigCodeBench50.0
Arena Elo1373.0

How to Run DeepSeek-V3 684.5B

Run DeepSeek-V3 684.5B locally with Ollama (needs 418.9 GB VRAM at Q4_K_M):

ollama run deepseek-v3