Back to AI intel
重点
vLLM Releases v0.30.0, Adds DeepSeek-V4.1-Flash Support with MXFP8 KV Storage
AI intel briefing
Core summary
One sentence to understand this update
`vLLM` released version `v0.30.0`, featuring 762 commits and adding support for new models including DeepSeek-V4.1-Flash, which utilizes MXFP8 for KV storage via FlashMLA V4.1.
Impact & opportunity
What this could mean
Developers can now leverage `vLLM` to efficiently deploy and serve DeepSeek-V4.1-Flash and other new models, benefiting from advanced memory optimization like MXFP8 KV storage.
Source
View original