Dessa AI is a new voice model from Block, the company behind Square and Cash App. Announced on 23 October 2026, it’s a full-duplex speech-to-speech model: it listens and talks at the same time, like a person on a phone call. What makes it unusual is how it was built. Block trained it from scratch on human conversation audio only, with no text and no pretrained models. In the video below I cover another big voice launch, Mistral’s Voxtral 3, so you can see where voice AI is heading. Here’s what Dessa is, how it works and what it means for you.
Key takeaways
- Dessa is Block’s full-duplex speech-to-speech AI.
- Trained only on human conversation audio — no text, no pretrained models.
- Block says it runs locally on a MacBook Pro M4 Max or better.
- Access is via a waitlist; AI safety researchers go first. Claim a free strategy session to plan voice AI for your business.
What is Dessa AI?
Most voice assistants work in steps. They turn your speech into text, send the text to a language model, then turn the answer back into speech. Dessa skips all of that. It’s a single model that takes in sound and puts out sound, and it does both at once. That’s what “full-duplex” means: it can hear you while it’s talking, so you can interrupt, laugh or say “mm-hmm” and it reacts the way a person would.
Block calls Dessa “the most authentic full-duplex speech model ever built”. Jack Dorsey says talking with it feels like talking with a person, full of emotion, nuance, pauses and a natural rhythm.
| Dessa at a glance | |
|---|---|
| Made by | A small AI research lab at Block (the company behind Square and Cash App) |
| Team | Founded by former members of the Toronto deep-learning startup Dessa; led by Ragavan Thurairatnam and reporting to Jack Dorsey |
| Announced | 23 October 2026 |
| Type | Full-duplex speech-to-speech conversation model |
| Trained on | Human conversation audio only — no text and no pretrained models |
| Training hardware | One NVIDIA GB200 NVL72 rack |
| Runs on | A MacBook Pro M4 Max or better, faster than real time (per Block) |
| Access | Public waitlist; early access for a small group of AI safety researchers |
How Dessa was trained
This is the headline. According to Block’s announcement, the team trained Dessa from scratch on nothing but human conversation. There was no text and no existing language model underneath. Dessa learned language from listening, the way a child does.
Block says Dessa learned from “hundreds of times fewer words” than large language models see. That’s why the pauses, laughs and back-and-forth sound so natural: those things are in real conversations, but they’re lost when speech is turned into text.
The training ran on a single NVIDIA GB200 NVL72 rack. That’s a lot of computing power for one company, but small compared with the huge clusters used for frontier text models.
Want to know how voice AI could answer your calls and win you leads? Claim your free strategy session →
Want AI that answers calls and books clients for you? Inside the AI Profit Boardroom you get the Agent OS — Hermes, OpenClaw, OpenCode, Claude and Codex in one place — plus step-by-step Hermes training and four live coaching calls a week.
Want a plan for your business first? Claim your free strategy session.
Join the AI Profit Boardroom →Claim Your Free SEO Strategy Session →
What runs Dessa: small enough for a laptop
Block says Dessa runs locally on a MacBook Pro M4 Max or better, and faster than real time. Jack Dorsey summed it up: it’s big news, but small enough to run on your MacBook. Ragavan Thurairatnam, who leads the team, says it runs on consumer GPUs.
Local voice AI matters for three reasons:
- Privacy — your conversations don’t have to leave your device.
- Speed — no round trip to a server means fewer awkward delays.
- Cost — no per-minute API bill once it’s running.
What Dessa sounds like
Block published several live recordings, including Dessa talking with a third-party model called GPT-Live-1. Block notes that GPT-Live-1’s developer isn’t affiliated with Block and hasn’t endorsed Dessa. Early listeners say it sounds remarkably human, and some find the realism a little eerie. That reaction is exactly why Block is starting with AI safety researchers.
Dessa AI vs other voice AI
| Dessa | Typical voice assistant | |
|---|---|---|
| How it works | One speech-to-speech model | Speech → text → model → text → speech |
| Listening while talking | Yes (full-duplex) | Usually takes turns |
| Trained on | Conversation audio only | Mostly text, plus a separate voice model |
| Pauses, laughs, “mm-hmm” | Learned from real speech | Often lost or added artificially |
| Runs locally | MacBook Pro M4 Max or better (per Block) | Usually in the cloud |
| Available now | Waitlist only | Widely available |
The trade-off is reach. Text-trained models know far more facts and can use tools. Dessa is built for natural conversation. Expect future voice agents to combine both: a Dessa-style voice in front of a capable agent like Hermes behind it.
How to get Dessa AI
You can’t download Dessa today. Block has opened a public waitlist at block.xyz/dessa, and early access is going to a small group of AI safety researchers who apply with an institutional email. Block hasn’t announced open weights, a licence, pricing or a public release date. See my Dessa waitlist guide for how to sign up.
Watch out for fake “Dessa” apps and downloads. The only official source right now is Block’s own page.
What Dessa AI means for your business
Natural voice AI will change how businesses handle phone calls, bookings, support and sales. Today most AI phone agents still sound robotic because of the speech-to-text step. Models like Dessa point to AI receptionists and sales assistants that people are happy to talk to.
- Answer every call and book appointments, even out of hours.
- Qualify leads with a natural conversation before handing over to your team.
- Run voice support that doesn’t frustrate customers.
- Keep data private by running voice models on your own hardware.
You don’t need to wait for Dessa to start. You can already set up voice agents today, as I show with Hermes Apollo voice.
My verdict on Dessa AI
Dessa shows a different path for AI: learn from speech, not text, and make it small enough to run on a laptop. It isn’t available to most people yet, but it’s a clear sign that natural, private voice agents are coming soon. Get on the waitlist and get your business ready to use voice AI, because Dessa AI won’t be the last model like this.
About Julian Goldie
I’m Julian Goldie — founder of Goldie Agency and the AI Profit Boardroom. I test new AI agents almost every day for 432K+ YouTube subscribers and run Hermes, OpenClaw, OpenCode, Claude and Codex side by side in my own Agent OS. Claim your free strategy session or join the AI Profit Boardroom.
FAQ: dessa ai
What is Dessa AI?
Dessa is a full-duplex speech-to-speech conversation model from Block, trained only on human conversation audio, with no text and no pretrained models.
Who made Dessa?
A small AI research lab at Block, founded by former members of the Toronto startup Dessa, led by Ragavan Thurairatnam and reporting to Jack Dorsey.
Can Dessa run on a laptop?
Block says it runs locally on a MacBook Pro M4 Max or better, faster than real time.
Is Dessa AI open source?
Block hasn’t announced open weights or a licence. Access is currently through a waitlist.
How do I try Dessa?
Join the waitlist at block.xyz/dessa. Early access is going to AI safety researchers first.
Ready to put Hermes to work in your business? Join the AI Profit Boardroom for the Agent OS, setup training and daily updates.
Or claim your free strategy session and we’ll map out your setup.
Starting from zero? Get the free AI course and community.
Join the AI Profit Boardroom →Claim Your Free SEO Strategy Session →Get the FREE AI Course + Community →
Related reading
Last updated October 2026. This page is a living guide to dessa ai — the facts here move fast and it is updated as they do.
