vLLM v0.30.0 ships a persistent GPU weight-cache daemon that skips the disk on restart.

Published September 22 with 762 commits from 315 contributors. Fast Start adds a per-GPU daemon holding post-quantized, TP-sharded weights in GPU memory, so restarting engines map them over CUDA IPC with --load-format ipc_cache, now covering FP4 and multi-node TP. The release also adds Gumbel-max watermarked generation with keyed-PRF detection, HiSparse host-resident KV spill for sparse-MLA decode, and DeepSeek-V4.1-Flash support. Freezing GC during CUDA graph capture cut capture from 12s to 2s and engine init from 28.9s to 8.2s on an H200. (vLLM)