New experiencesFlagship

On-Device AI Assistant

One assistant across vehicle, phone, and store. Voice, vision, and language run where the user is; nothing leaves the device, and there's no per-query bill.

Pick a stage

talk trackThe whole voice turn happens on the head unit: speech in, spoken answer out. Try it live, this one is running right now.

Speech in, speech out: one model, no pipeline

LFM2.5-Audio-1.5B handles the whole voice turn natively — recognition, multi-intent understanding, and the spoken reply — while a 1.2B tool-calling model drives 100+ vehicle functions. "Turn on heated seats and navigate home" resolves both intents in one pass, entirely on the head unit. No connectivity, no per-query cost, no audio leaving the cabin.

Open the live voice cockpit demo →
  • End-to-end audio: no separate STT → LLM → TTS chain
  • Native tool-call format drives climate, media, navigation
  • Multi-turn context — follow-ups without restating
  • 100% offline; audio never leaves the cabin (the hard part of GDPR/PIPL/CCPA)