vllm - 💡(How to fix) Fix [Usage]: DeepSeek-V4 startup takes ~8 min — guidance on reducing initialization time

Your current environment

node: h20

How would you like to use vllm

node: h20

When serving DeepSeek-V4-Flash with the following command, the time from process start to "service ready" is approximately 8 minutes. I would like to understand whether this is expected and what are the recommended ways to speed up the startup phase. Do you have any documentation on startup speed optimization?

vllm serve /models/deepseek-ai/DeepSeek-V4-Flash \
  --trust-remote-code \
  --kv-cache-dtype fp8 \
  --block-size 256 \
  --enable-expert-parallel \
  --tensor-parallel-size 8 \
  --tokenizer-mode deepseek_v4 \
  --tool-call-parser deepseek_v4 \
  --enable-auto-tool-choice \
  --reasoning-parser deepseek_v4

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.