Block's Dessa is a speech-to-speech model trained on 'nothing but human conversation' — full-duplex, running locally on 'a MacBook Pro M4 Max or better', trained 'on only one NVIDIA GB200 NVL72' — waitlist and safety-researcher early access only, no public availability yet

“Block built a voice AI that learned to talk the way you did — by listening, with no text at all — and it runs locally on a MacBook Pro M4 Max.”

Usable claims

• On Friday, October 23, 2026 (calendar-verified), Block announced Dessa, a speech-to-speech voice model from its internal AI research lab, via the model's own page at block.xyz/dessa — a page that carries no date label of its own (the request-access form's Privacy Notice 'Effective Date: October 21, 2026' is embedded form chrome), dated October 23 by Block's own announcement posts on X, whose date labels read that day (the @blocks post shows '6:50 PM · Oct 23, 2026'). Per the page: 'We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.' The training claim: 'We trained Dessa from scratch on nothing but human conversation. Sounds! No text, no pretrained models. It learned to hold a real-time conversation simply by listening.' — with Block's own hedge on the first-of-kind claim: 'As far as we know, Dessa is the world's first AI you can have a realistic live conversation with that was trained only from human speech.' The mechanics: 'Audio in, audio out. Dessa is a true speech-to-speech model, and everything it knows came from human speech. It's full-duplex, so it listens and talks at the same time, like you do.'; 'It learned from real conversation, so it knows when to cut in and when to backchannel ("yeah!", "got it!", "ooh") to show it's with you.'; and 'Inference runs faster than real time, so there's no awkward wait. And it runs locally on a MacBook Pro M4 Max or better.' The emotion claim: 'It natively hears and speaks emotion, sarcasm, laughter, a frustrated sigh, a voice cracking with sadness.' Block's data-efficiency framing, its own words: 'LLMs need an insane amount of text. Dessa learned language from hundreds of times fewer words (in fairness, LLMs also have to learn C++ and whatever the hell Angular is).' The compute line: 'we figured out how to train Dessa on only one NVIDIA GB200 NVL72' (the page bolds 'one'; a GB200 NVL72 is a rack-scale system — TechCrunch's rendering 'a single NVIDIA GPU' is the outlet's compression, kept per-source and never merged). The samples: 'Every clip below is a live recording of a person talking to Dessa. Nobody steers it. Dessa picks what to say, when to say it and how, down to its own voice and personality. Subtitles were added afterward.' — including a demo where 'Dessa is talking with GPT-Live-1, a third-party AI model. Both voices are AI-generated. The developer of GPT-Live-1 is not affiliated with Block and has not reviewed or endorsed Dessa.' The lab: 'We're a (very) small AI research lab at Block, founded by former members of Dessa, the Toronto deep learning startup. We've built deep learning models since 2012 and generative speech models for over eight years.' and 'we report directly to Jack Dorsey, Block's cofounder, Chairman and Block Head'. Availability: the page offers 'Join the waitlist' plus a request-access application form, and per the page's 'Calling all safety researchers!' section — 'We're giving a small group of AI safety researchers early access to find the risks we haven't anticipated.' signed '— Ragavan Thurairatnam, Dessa lead' — while 'More recordings will be shared on X soon.' is the announcement body's only forward-timing line (the embedded privacy-notice chrome separately reads 'Launch links are coming soon.'). Per TechCrunch (Ivan Mehta, published_time 2026-10-24T16:00:00+00:00): 'Jack Dorsey's Block released a new voice model out of the company's research lab that runs locally and sounds strangely human with its expressions like laughs, pauses, and cries.' with the outlet's availability read: 'However, there is no public access to Dessa at the moment, and Block has put out a call for safety researchers to test the model.' and its business context: 'Square already uses voice AI to handle ordering.'