Which AI model should run your Home Assistant voice assistant?
Accuracy at controlling a smart home, from published Home Assistant eval results. Speed & latency on real hardware coming from our own runs.
| # | Model | Type | Overall | Assist (n=460) | Assist-mini (n=196) | Questions (n=370) | Tokens/sec | Response latency |
|---|
Get the weekly "new model tested" email
One short email when a new local model is benchmarked for Home Assistant voice. (Placeholder — signup not connected yet.)