Block announced Dessa on 23 October through Ragavan Thurairatnam, who leads its in-house AI research lab. It is a full-duplex speech-to-speech model trained from scratch on recorded human conversation alone - no text, no pretrained models. Because it takes audio in and returns audio out, it never converts speech to text, which Block argues is where voice systems lose the information carried in sound itself. Full-duplex means it can listen and speak at the same time rather than waiting for a turn to end. Block says it learned from hundreds of times fewer words than a large language model.
The hardware numbers are the part worth noting. Dessa was trained on a single NVIDIA GB200 NVL72 system and runs locally on a MacBook Pro M4 Max or better. The lab behind it names twelve people - a lead, six researchers, three interns and two former interns - and was founded by former members of Dessa, the Toronto deep learning startup Block acquired. It reports directly to Jack Dorsey.
Access is deliberately narrow. Early access is limited to a small group of AI safety researchers who must apply with an institutional email address, with a public waitlist open alongside it. No pricing has been announced, and no launch date beyond "coming soon".
Sources
Unite.AI · CryptoBriefing