Gizmodo reported the announcement on October 24, 2026 and published examples of the system speaking in different voices and emotional registers.
Dessa uses a speech-to-speech, duplex architecture: according to the report, it processes and generates audio directly rather than first transcribing speech to text. That design lets it listen and speak at the same time. Block says the model can run locally on a MacBook Pro with an M4 Max chip or better.
Access is not generally available. Block says it has given early access to a small group of AI safety researchers. The report does not establish a public release date, API or application access, pricing, model size, training-data details, benchmark results, or independent evaluation of its voice quality and safety.