Rent a GPU

DeepSeek V4 Flash 284B-A13B hardware requirements

284B parameters (13B active, mixture of experts). Needs about 173 GB at Q4_K_M with 8K context.

Parameters284B13B active (MoE)
Q4_K_M · 8K173 GBthe usual download
FP16 · 8K568.8 GBfull precision
Runs well on1/51GPUs and Macs we track

Memory needed at each quantization

Weights, KV cache and runtime overhead, at 8K and 32K tokens of context.

QuantQuality8K context32K context
FP16Full quality568.8 GB569.4 GB
Q8_0Near-lossless302.6 GB303.2 GB
Q6_KExcellent233.7 GB234.3 GB
Q5_K_MVery good202.8 GB203.4 GB
Q4_K_MGood, most popular173 GB173.6 GB
Q3_K_MNoticeably degraded139.6 GB140.2 GB
Q2_KHeavily degraded107.3 GB107.9 GB

Run it with Ollama

Downloads Q4_K_M by default.

ollama run deepseek-v4-flash

Model card

deepseek-ai/DeepSeek-V4-Flash-0731 on Hugging Face: licence, prompt format and GGUF builds.

Run it in the cloud

Run DeepSeek V4 Flash 284B-A13B on a rented GPU today.

It needs about 173 GB at Q4_K_M: a rented 4× A100 / H100 runs it for about $4.44 to $15.96 an hour.

RunPodRent a GPU on RunPodVast.aiRent a GPU on Vast.ai

Every GPU and Mac checked

Sorted by memory. Bars show how much of each card DeepSeek V4 Flash 284B-A13B fills at the best quant that fits.

1 Runs well

NameMemory usedBest quantSpeed

1 Fits with heavy quantization

NameMemory usedBest quantSpeed

49 Too big

NameMemory usedBest quantSpeed

Other DeepSeek models