SGLang

DeepSeek-V4 is introduced with features like fast inference and verified reinforcement learning (RL) using SGLang and Miles. The article highlights its launch on LMSYS.org, emphasizing advancements in AI model efficiency and safety.

A critical security vulnerability, CVE-2026-5760, in SGLang allows remote code execution via malicious GGUF model files. The flaw, with a CVSS score of 9.8, stems from command injection and affects the high-performance open-source serving framework.