Woosuk Kwon(@woosuk_k):DeepSeek is so back! @deepseek_ai keeps pushing the limits of KV cache compression with new architectural ideas. Huge respect. 🫡 And of course, vLLM has full day-0 support!

2026.09.10 07:24

DeepSeek is so back! @deepseek_ai keeps pushing the limits of KV cache compression with new architectural ideas. Huge respect. 🫡 And of course, vLLM has full day-0 support!

vLLM@vllm_project

2026.09.10 06:50

🐳 DeepSeek-V4.1-Flash is out, and vLLM serves it from day 0, verified on NVIDIA and AMD GPUs! 🎉 552B MoE backbone, native vision, 1M context. Built for agents: 8B active while it reads your prompt, 16B while it writes. If you already run DeepSeek-V4 on vLLM, most of this stack will feel familiar: the hyper-connections, the sliding-window plus compressed sparse attention, DSpark drafting, MXFP4 experts. vLLM has carried all of it since V4 landed. Two things are new, and both are worth a look: ✨ Engram: a quarter of the checkpoint is n-gram memory the model looks up instead of computes. 197B parameters of it. ✨ Only four layers write compressed KV now. The rest of the model shares it. Spin it up 👇 🔗