LLMDeepSeekMIT

DeepSeek V4 Pro

deepseek-ai/DeepSeek-V4-Pro

Parameters

1600B

Context

1024K

Downloads

1.4M

Likes

5.4K

Architecture

moe_transformer

View on HuggingFace

GPU Requirements

VRAM / RAM / disk estimates by quantization and usage scenario

QuantizationMinimumRecommendedProduction
FP1616bit
3774.9 GB
RAM 5662.4 · Disk 4688.2
3962.1 GB
RAM 5943.1 · Disk 4714.2
4230.3 GB
RAM 6345.5 · Disk 4753.2
FP88bit
1890.0 GB
RAM 2834.9 · Disk 2363.6
1991.4 GB
RAM 2987.1 · Disk 2389.6
2174.0 GB
RAM 3260.9 · Disk 2428.6
INT88bit
1890.0 GB
RAM 2834.9 · Disk 2363.6
1991.4 GB
RAM 2987.1 · Disk 2389.6
2174.0 GB
RAM 3260.9 · Disk 2428.6
AWQ4bit
982.8 GB
RAM 1474.2 · Disk 1244.9
1043.0 GB
RAM 1564.5 · Disk 1270.9
1184.3 GB
RAM 1776.5 · Disk 1309.9
GPTQ4bit
994.6 GB
RAM 1491.9 · Disk 1259.4
1055.3 GB
RAM 1583.0 · Disk 1285.4
1197.2 GB
RAM 1795.8 · Disk 1324.4
GGUF4bit
1018.1 GB
RAM 1527.2 · Disk 1288.5
1079.9 GB
RAM 1619.9 · Disk 1314.5
1222.9 GB
RAM 1834.3 · Disk 1353.5

GPU Recommendations

Best ValueBEST

8× H200 SXM

1128 GB VRAM

$20,440/mo

≈ 19.9 tok/s

Effective production throughput ≈ 13.9 tok/s

score 0.3731

Cheapest

8× H200 SXM

1128 GB VRAM

$20,440/mo

≈ 19.9 tok/s

Effective production throughput ≈ 13.9 tok/s

score 0.3731

Performance

8× H200 SXM

1128 GB VRAM

$20,440/mo

≈ 19.9 tok/s

Effective production throughput ≈ 13.9 tok/s

score 0.3731

Min Complexity

8× H200 SXM

1128 GB VRAM

$20,440/mo

≈ 19.9 tok/s

Effective production throughput ≈ 13.9 tok/s

score 0.3731

💬 Community — real deployment experience

0 Articles · 0 Benchmarks

View Community →
AmciHub

AI Model Deployment & Compute Intelligence — analyze AI model requirements, GPU performance, cloud pricing and deployment costs to find the right deployment solution.

© 2026 AmciHub. All rights reserved. Model → Requirement → GPU → Benchmark → Cloud → Cost → Recommendation