Suitably MS is behind a bunch of small models released to hugging face for things like audio transcription and text to speech. Suggests they’re moving to the idea of full local experience, but haven’t really been able to fully materialise it.

Wouldn’t surprise me if they’re fighting it out internally of having the fully local experience they’re kinda moving to vs the frontier experience using cloud and subscriptions they can change more for.

To be honest, Microsoft’s software is running against an API that uses a translation layer to talk to an older API. Multiple layers of this happen before you get to hardware, mostly for compatibility with legacy software that uses the older APIs.

Not to mention MS have rewritten their frontends for their apps several times and keep them each time.