MODEL · GPU GUIDE

DeepSeek-R1-0528-NVFP4-v2 GPU Requirements: VRAM & Cheapest GPU

DeepSeek-R1-0528-NVFP4-v2 has about 394B parameters. See exactly how much GPU memory it needs at FP16, INT8, and INT4, and the cheapest GPU to run it, with live hourly pricing from 5+ data center partners.

394BParameters
215 GBMin VRAM
$3.28/hrCheapest
< 2 minDeploy
nvidia/DeepSeek-R1-0528-NVFP4-v2
VIEW ON HUGGINGFACE ↗
394B paramstext-generationdeepseek_v31.4M downloads23 likesupdated Sep 2, 2025

To run DeepSeek-R1-0528-NVFP4-v2 for inference at FP16, you need roughly 858 GB of VRAM. The cheapest fit on Spheron is 8x H200 141GB at about $26.48/hr. Quantize to INT4 to run it on a smaller, cheaper GPU.

GB VRAM REQUIRED
FP16INFERENCEBATCH 1CTX 4k

Estimated peak VRAM including weights, activations, and KV cache. Add 10% headroom for production traffic.

RANKCONFIGURATIONPER GPUTOTAL $/HR
  • 01
    8× H200 141GBCHEAPEST
    Hopper · HBM3e
    $3.31/hr$26.48/hr
  • 02
    4× B300 288GB
    Blackwell Ultra · HBM3e
    $10.17/hr$40.68/hr
  • 03
    8× B200 192GB
    Blackwell · HBM3e
    $5.34/hr$42.72/hr

Live pricing aggregated from 5+ data center partners. Per-minute billing, no commitments.

VRAM required to run DeepSeek-R1-0528-NVFP4-v2

Estimated peak VRAM at context length 4,096 and batch size 1, including weights, activations, and KV cache. Quantizing to INT8 (Q8) or INT4 (Q4) cuts memory roughly in half and in quarter.

PrecisionInferenceLoRA fine-tuneFull fine-tune
FP16858 GB1287 GB3432 GB
INT8429 GB644 GB1716 GB
INT4215 GB322 GB858 GB

Cheapest GPU to run DeepSeek-R1-0528-NVFP4-v2 by precision

FP16
VRAM required858GB

Full precision. Best quality, highest memory.

Cheapest GPU
8x H200 141GB
Hopper · HBM3e
$26.48/hr · $3.31/hr/gpu
8x H200 141GB on Spheron
INT8
VRAM required429GB

8-bit quantized. ~2x smaller, minimal quality loss.

Cheapest GPU
8x A100 80GB
Ampere · HBM2e
$6.56/hr · $0.82/hr/gpu
8x A100 80GB on Spheron
INT4
VRAM required215GB

4-bit quantized. ~4x smaller, runs on smaller GPUs.

Cheapest GPU
4x A100 80GB
Ampere · HBM2e
$3.28/hr · $0.82/hr/gpu
4x A100 80GB on Spheron

Inference vs fine-tuning DeepSeek-R1-0528-NVFP4-v2

InferenceWeights + KV cache
LoRA fine-tune~1.5×+ low-rank adapter
Full fine-tune~4×+ gradients + optimizer state

Inference only holds the model weights plus a KV cache, so it is the cheapest setup. LoRA fine-tuning adds a small adapter and roughly 50% more memory. Full fine-tuning holds gradients and optimizer state on top of the weights, which is about 4x the inference footprint, so it often needs multiple GPUs even when inference fits on one. For DeepSeek-R1-0528-NVFP4-v2, an on-demand H200 141GB instance covers inference and LoRA, while a full fine-tune needs several times that memory and often spans multiple GPUs. Check the live GPU pricing for current rates.

Similar models

Compare GPU requirements for models in the same class.

FAQ / 05

DeepSeek-R1-0528-NVFP4-v2 GPU questions