Xunzhuo(@XunzhuoLiu):No single decision model wins. Introducing System One Auto in vLLM Semantic Router. Kai 0.6B answers first; Vega 27B only when needed. Public JevBench requests: -54% est. cost, -20% mean latency vs Vega-only +16.45 pts accuracy vs Kai alone #Jev #vLLM

Xunzhuo

@XunzhuoLiu

Founder of | Focus on Intelligent Routing @vllm_project | Building Foundation Decision Models | Open Source and AI @AMD | Think different

加入 January 2026

20 正在關注 941 粉絲

Xunzhuo@XunzhuoLiu

2026.10.10 10:17

No single decision model wins. Introducing System One Auto in vLLM Semantic Router. Kai 0.6B answers first; Vega 27B only when needed. Public JevBench requests: -54% est. cost, -20% mean latency vs Vega-only +16.45 pts accuracy vs Kai alone #Jev# #vLLM#