Welcome to vLLM office hours! These bi-weekly sessions are your chance to stay current with the vLLM ecosystem, ask questions, and hear directly from contributors and power users. This week’s special topics: What’s New in Speculators & Accelerating Sparse MLA We cover the latest vLLM project updates, including new model support, Model Runner V2 progress, Rust frontend work, KV cache offloading, Blackwell RTX improvements, and more. We also introduce the first engineering spotlight with a walkthrough of masked MHA for accelerating sparse MLA. Then we dive into the latest Speculators release, covering speculative decoding fundamentals, response regeneration, Eagle 3, P-Eagle, DFlash, MTP fine-tuning, the new evaluation suite, improved documentation, hidden state extraction with vLLM, and a live demo training and serving a speculative decoding model during Office Hours. Slides: https://docs.google.com/presentation/d/146fadQKacCEBnVmCRwPPFS-OXu34yoyvDGCVyWjhg7o/ Want to join the discussion live on Google Meet? Get a calendar invite by filling out this form: https://red.ht/office-hours Timestamps: 00:00 vLLM intro 03:13 vLLM v0.22 project update 18:26 Engineering spotlight: masked MHA for sparse MLA 27:06 Speculators update 55:18 Q&A

 341      7