vLLM 0.25.0 Released
mervenoyann · x · 2026-07-12
vLLM has released v0.25.0, a massive update comprising 558 commits and 232 contributors, including 64 new contributors.
Key changes include:
- Model Runner V2 becomes the default for all dense models
- Legacy PagedAttention implementation removed
- Transformers backend is now as fast as native vLLM
- New unified Streaming Parser Engine
- Universal speculative decoding (TLI) across heterogeneous vocabularies
- New DSpark and DFlash drafters
- Support for new models Hy3 and Unlimited OCR
This is a major low-level and engineering update, focusing on inference stack performance and architectural consolidation.
More from Infra
- SF Compute founder: buying compute is 'an absolutely awful experience' right now — IgorCarron · 2026-09-11
- SmolVM open-sources persistent computer infrastructure for agents that outlive chat sessions — aniketmaurya · 2026-09-11
- PyTorch Day Korea 2026 launches first offline conf, CFP closes Sept 13 — PyTorch · 2026-09-11
- Local LLM server dilemma: 4x CMP-170HX (price up 53% in 20 days) vs Mac Studio M5 Ultra — rumboll · 2026-09-11
- llama.cpp lands Flash Attention tuning for RDNA4, big prefill gains on AMD — pmttyji · 2026-09-11
- Your p99 latency benchmark may be lying: a deep dive into coordinated omission — Franc0Fernand0 · 2026-09-11