google-ai-edge/litert-lm (LiteRT-LM) is Google's open-source inference framework, a project for deploying large language models (LLMs) on edge devices. The repository description states that the framework is in a shape usable for production. The latest release of this repository is v0.18.0, and the source is the google-ai-edge/litert-lm repository on GitHub.
LiteRT-LM is Google's orchestration layer (an intermediate layer that manages the execution flow) that runs LLMs across multiple platforms using LiteRT. The README explains that it supports Android, iOS, web, desktop, and IoT, and provides GPU/NPU acceleration, vision and audio input, and function calling. Supporting local language, vision, and speech inference on Android, iOS, IoT, and Raspberry Pi also falls within this scope. The implementation language is C++ and the license is Apache-2.0. The GitHub star count is 6700.
However, the basis for a judgment is thin. The given record contains no release notes for v0.18.0, so what changed in this version is unknown. There are no performance benchmarks and no per-model runtime or memory figures, and the score only reaches a momentum of 2.8 out of 10, with the novelty, technical, and quality items empty. The repository was created on April 14, 2025, so its history is not long. Until the release notes and actual deployment cases are checked directly, it is hard to cite as grounds for adoption.