AI processing architecture
Device, phone, local-host and cloud processing across documented product configurations.
Processing stages
An AI device can capture input locally and use another computer for inference, response generation or storage. Architecture comparisons require documentation for each stage and its connection requirements.
Local hosts and endpoints
Home Assistant Voice Preview Edition is an endpoint. Its voice pipeline can use a separate local host or cloud services. Local operation of the complete setup does not establish that the endpoint itself runs a language model.
Feature-level routing
Brilliant Labs describes an NPU inside Halo, models on the phone and a cloud-based agent. The reviewed documentation leaves some feature-level routing and offline-operation requirements unresolved.
Privacy and network dependencies
Local processing capability and privacy are separate properties. A local-capable computer can run software that contacts external services. Privacy assessment requires data-flow documentation for the specific feature and configuration.
This is our interpretation of the linked documentation. No hands-on testing or independent benchmarking was performed.
Primary sources
Sources checked November 4, 2026. Product claims and documentation can change.