Back to the Glossary

Running AI models directly on the devices themselves, such as phones, cameras and cars, instead of sending data to the cloud. It gives near-instant responses, keeps data on the device and works offline; the trade-off is limited compute and memory, which means smaller models. It fits when speed or privacy matters more than raw model power, as in phones that run AI on a built-in neural processing unit (NPU).

More in the same category

Browse the full glossary