Llama 1B running on the glasses themselves

Qualcomm demonstrated the AR1+ chip running Meta's Llama 1B model fully on-device at AWE 2025, with no phone or cloud required, meaning some Gen 3 AI queries could run locally instead of routing through a paired phone if the glasses end up using that chip, which hasn't been confirmed.

Where the AI runs is the whole point of the AR1+ demo. Qualcomm showed prototype glasses running Meta's Llama-3.2-1B model fully on-device, with no phone and no cloud. If Gen 3 uses that chip, some AI queries would answer on the frames instead of routing through a paired phone, which cuts the lag and works with your phone still in your pocket. The chip-in-Gen-3 pairing is not confirmed, so this hangs on that.