LM Studio 0.4.0 adds headless serving and concurrent inference

Release and source context

Element Labs released LM Studio 0.4.0 with the llmster headless core, concurrent requests and continuous batching in its llama.cpp engine, and a stateful chat API. The release expanded local model serving beyond the desktop interface. [1]

The January 28 launch introduced llmster headless operation, concurrent requests, continuous batching and a stateful chat REST API. Later build notes do not describe only launch-day features. Engine 2.0.0 is LM Studio's packaging of its llama.cpp engine, rather than an upstream llama.cpp major release. [1]

How-To Geek published its contemporary report on January 28, 2026. [2]