Last released Oct 23, 2026
A high-throughput and memory-efficient inference and serving engine for LLMs