Back to AI intel
重点
搞钱
vLLM v0.30.0 Release Features DeepSeek-V4.1-Flash and MXFP8 KV Storage
AI intel briefing
Core summary
One sentence to understand this update
vLLM's v0.30.0 release introduces support for new models like DeepSeek-V4.1-Flash and optimizes KV storage using MXFP8 on SM100, benefiting from 762 commits and 315 contributors.
Impact & opportunity
What this could mean
Developers can leverage the new models and optimized memory usage in vLLM for more efficient and powerful large language model deployments.
Source
View original