// event · library

vLLM 0.30.0 adds DeepSeek-V4.1-Flash, IPC weight-cache fast-start, watermarking, HiSparse KV offload

library

vendor: vllm-projectpublished: 21d agoconfidence: 0.95id: 221c3cd3-00b5-431d-a4eb-cb8cbd618af6

vLLM 0.30.0 lands 762 commits: new model support for DeepSeek-V4.1-Flash, DeepSeek-V4-Flash-Vision-Exp, GLM-5.3-Flash, K2-Horizon, Cohere Compass and Bailing V3 VL; a persistent IPC weight-cache daemon (`--load-format ipc_cache`) for fast engine restarts; Gumbel-max watermarking with detection; HiSparse host-tier KV offload; Model Runner V2 dual-batch overlap; PCP+DCP large-scale serving; and NVFP4/AutoRound quantization upgrades.

capability delta

an agent can now serve DeepSeek-V4.1-Flash, GLM-5.3-Flash, K2-Horizon and Kimi K3 with watermarked generation, HiSparse host-tier KV offload, IPC weight-cache fast-restart, and PCP+DCP large-scale serving via vLLM 0.30.0.

// install

pip install vllm==0.30.0

pypi · vllm @ 0.30.0

// example

bash

pip install vllm==0.30.0

# Fast restart via persistent per-GPU weight cache daemon
vllm serve deepseek-ai/DeepSeek-V4.1-Flash \
  --load-format ipc_cache \
  --kv-cache-dtype fp8_ds_mla \
  --enable-scale-out

// links

vllminferenceservingdeepseekglmkimiquantizationkv-cachewatermarkingspeculative-decoding

source: github_release · https://github.com/vllm-project/vllm/releases/tag/v0.30.0