← Back to feed

LiteRT-LM

GitHub Repo Pretty sure · Google ships it; that matters.
https://github.com/google-ai-edge/LiteRT-LM

Google's edge LLM runtime that actually runs on devices instead of just talking about it—shipping in Chrome, Pixel Watch, and production Android apps with real performance gains.

15%
60%
25%
Slop 15%Signal 60%Science 25%

LiteRT-LM is a legitimate inference runtime, not marketing. It powers actual Google products (Chrome, Pixel Watch, Chromebook) and solves the real problem: running LLMs on constrained hardware without network dependency. The signal is strong—quantized Gemma models running on Raspberry Pi with measurable latency/throughput is useful. Science score modest because this is engineering execution, not novel research (quantization, KV-cache optimization, hardware scheduling are solved problems). Slo...

2510 stars C++ 2026-04-07 358 days old

Become a MFer to rate — log in