LocalForge
Voltar ao guia
MODEL / DEEPSEEK

DeepSeek V4 Flash

The 176 GB memory figure is an estimate for the mixed-precision checkpoint. Thirteen billion active parameters do not mean 13 GB of storage.

Fonte primária
Parâmetros totais284B13B active / token
Memória estimada~176 GBOfficial FP4 + FP8 mixed
Contexto1M nativeText
LicenseDeepSeek Model LicenseMultilingual
01 / IDEAL PARA
  • long-context agents
  • coding
  • reasoning
02 / AJUSTE ESTIMADO
Mac Studio M3 Ultra · 256 GBTight
03 / RODAR COM
vLLMAdvancedvllm serve deepseek-ai/DeepSeek-V4-Flash --dtype auto
llama.cppMediumllama-server -m /path/to/verified-model.gguf --port 8080
LM StudioEasySearch “deepseek-ai/DeepSeek-V4-Flash” → Load a Q4/INT4 build → Start Server
Atenção

The 176 GB memory figure is an estimate for the mixed-precision checkpoint. Thirteen billion active parameters do not mean 13 GB of storage.