BIG NEWS! 🎉 Compressed Tensors is officially joining the vLLM! Built on top of the excellent Hugging Face safetensors framework, Compressed Tensors extends it with efficient storage and management of compressed tensor data for model quantization and sparsity. Why

💡 vLLM @ Open Source AI Week！ 1⃣ Wednesday, Oct 23 & Thursday, Oct 24: vLLM @ Pytorch Conference 2025 🚀 Explore vLLM at PyTorch Conference 2025! 📅 Sessions to catch: 1. Easy, Fast, Cheap LLM Serving for Everyone – Simon Mo, Room 2004/2006 2. Open Source Post-Training Stack:

We are excited about an open ABI and FFI for ML Systems from Tianqi Chen. In our experience with vLLM, such interop layer is definitely needed!

kvcached works directly with vLLM and you can use it to serve multiple models on the same GPU. They will share unused KV cache blocks. Check it out!

🚀 Excited to share our work on batch-invariant inference in vLLM! Now you can get identical results regardless of batch size with just one flag: VLLM_BATCH_INVARIANT=1 No more subtle differences between bs=1 and bs=N (including prefill!). Let's dive into how we built this 🧵👇

Catch us in the special vLLM track at Anyscale Ray Summit 2025 SF Marriott Marquis vLLM Deep Dive: Architecture, Performance & Contributing by Tan TJian Can’t wait to see all vLLM friends. anyscale.com/ray-summit/202…

vLLM Sleep Mode 😴→ ⚡Zero-reload model switching for multi-model serving. Benchmarks: 18–200× faster switches and 61–88% faster first inference vs cold starts. Explanation Blog by EmbeddedLLM 👇 Why it’s fast: we keep the process alive, preserving the allocator, CUDA graphs,

🔥 Following our big announcement — here’s the full vLLM takeover at Ray Summit 2025! 📍 San Francisco • Nov 3–5 • Hosted by Anyscale Get ready for deep dives into high-performance inference, unified backends, prefix caching, MoE serving, and large-scale

🎉 Congrats to Kimi.ai! vLLM Day-0 model expands! Now supporting Kimi Linear — hybrid linear attention with Kimi Delta Attention(KDA): - RULER 128k context: 84.3 perf + 3.98× speedup - Up to 6× faster decoding & 6.3× faster TPOT (1M tokens) - 75% KV cache reduction 💡

There are actually quite a few optimizations specific to multimodal other than just faster kernels, and I didn’t pull those all nighters for nothing🥲

Wow excited to see PewDiePie using vLLM to serve language models locally 😃 vLLM brings easy, fast, and cheap LLM serving for everyone 🥰

Wow Quantization-enhanced Reinforcement Learning using vLLM! Great job by Yukang Chen 😃

🔥Highly requested by the community, PaddleOCR-VL is now officially supported on vLLM! 🚀 Check out our recipe for this model to get started!👇docs.vllm.ai/projects/recip…

Amazing work by Rui-Jie (Ridger) Zhu and the ByteDance Seed team — Scaling Latent Reasoning via Looped LMs introduces looped reasoning as a new scaling dimension. 🔥 The Ouro model is now runnable on vLLM (nightly version) — bringing efficient inference to this new paradigm of latent

Happy to meet with EmbeddedLLM people in person! Thanks for all the hardware supports in vLLM and LMCache Lab !

Awesome connecting with the vLLM community in person! Roger Wang Kuntai Du Harry Mellor Simon Mo Tan TJian Chendi Xue Robert Shaw Brittany Rockwell

สวัสดีครับ Sawadekap, Bangkok! พร้อมจะโกลว์กันหรือยัง? ✨ vLLM Meetup — 21 Nov 2025 Hosted by EmbeddedLLM, AMD & Red Hat Members from the vLLM maintainer team will join us to share their latest insights and roadmap — straight from the source! We've also invited local Thai

Thanks to GitHub for spotlighting vLLM in the Octoverse 2025 report — one of the fastest-growing open-source AI projects this year. 🏆 Top OSS by contributors 🚀 Fastest-growing by contributors 🌱 Attracting the most first-time contributors Trusted by leading open model

Big night at the vLLM × Meta × AMD meetup in Palo Alto 💥 So fun hanging out IRL with fellow vLLM Woosuk Kwon, Simon Mo and the AMD crew Anush Elangovan and Ramine Roane. Bonus: heading home with a signed AMD Radeon PRO AI Pro R9700 to squeeze even more tokens/sec out of AMD