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 topic: GuideLLM v0.7.0: Benchmarking Reasoning, Tool Calling, and Agentic Traces. vLLM project update from core maintainer Michael Goin, covering the v0.27 release plus day-zero support for Qwen3.8, decode context parallelism, and adaptive verification for speculative decoding. Lightning topic 1: Running Codex and Claude Code CLIs with vLLM and open models, with Francisco Javier Arceo (Red Hat AI) and Tun Jian Tan (Embedded LLM). A look at the Agentic API project, which keeps stateful logic outside vLLM so the inference engine stays stateless, plus a live demo of server-side web search. Lightning topic 2: Speculators update from maintainer Rahul Tuli, covering the v0.6.0 and v0.7.0 releases, new algorithms including P-EAGLE and DSpark, MTP fine-tuning, and a config-file-first training CLI. Then our special topic, with Samuel Monson and Jared O'Connell (Red Hat AI): what performance benchmarking actually measures, how GuideLLM simulates realistic request patterns, and how the tool is evolving to handle multi-turn conversations, tool calling, sub-agents, and trace replay. Slides: https://docs.google.com/presentation/d/1GdxjSmHvwM-X6sylRrbHo9-0UYQOSpJMNvhxqFUXZSY 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 Intro 03:06 What's new in the vLLM community over the past two weeks 12:02 vLLM v0.27 release 15:50 Lightning topic: Codex and Claude Code CLIs with vLLM 22:40 Lightning topic: Speculators update 33:33 GuideLLM v0.7.0 deep dive 50:45 Q&A

 1840      14