Developer
Dessa AI API: What Block Has (and Hasn't) Announced for Developers
Developers are already searching for a Dessa AI API — our own search data shows the queries landing on this site — so here is the short, honest answer up front: Block has not announced a Dessa API. There is no documentation, no API key, no developer console, no pricing per minute and no SDK. Anyone selling you Dessa API access today is not Block. What follows is what Block has actually said, what a Dessa API would have to answer if one ever arrives, and what developers can use in the meantime.
What Block has announced — and what it has not
Block announced Dessa on October 23, 2026 as a research preview: a full-duplex, speech-to-speech model trained from scratch on human conversation audio only, with no text and no pretrained models, as we explain in what Dessa AI is. The announcement centred on demos, a public waitlist, and early access for a small group of AI safety researchers.
The list of things not announced is unusually clear, and developers should plan around it: no API, no weights, no download, no parameter count, no benchmarks, no pricing and no public release date. Our Dessa Tracker keeps that status board current. A research preview is not a developer platform, and Block has given no timeline for it becoming one.
Why the API question is different for Dessa
For most voice AI, an API is straightforward to imagine: send audio or text, get audio back, pay per minute. Dessa complicates that picture in two interesting ways. First, Block says inference runs faster than real time locally on a MacBook Pro with an M4 Max chip or better — a claim we explored in why Dessa runs on a laptop. A model designed to run on a user's own hardware points just as naturally toward an on-device release as toward a hosted API, and Block has announced neither.
Second, Dessa's defining behaviour is live, full-duplex conversation: listening while speaking, backchannelling, handling interruptions in real time. A developer API for that would not be a simple request-and-response endpoint; it would need a streaming, low-latency session model. That is buildable — rivals offer realtime voice APIs today — but it is a product decision Block has not made publicly, and the safety-first rollout suggests it will not be rushed. A voice this convincing raises impersonation questions that a public API would multiply, which is why safety researchers are first in line, as our waitlist guide explains.
What developers can use today instead
If you need voice in a product now, the honest alternatives are the systems that actually ship developer access. OpenAI offers realtime voice to developers at published rates. Google's Gemini Live is available through Google's developer surfaces. Kyutai's Moshi, which we compared in Dessa vs Moshi, released open weights, so researchers can run and inspect a full-duplex system themselves — something nobody can do with Dessa today, because Block has released no weights.
Building on those platforms now is not a wasted bet against Dessa. The interface lessons transfer: streaming audio, interruption handling, session state, latency budgets and disclosure design will matter for any speech-native model that eventually opens up. What does not transfer is any assumption about Dessa's pricing, terms or availability — none exist.
The signals that would change this answer
Four announcements would turn this article from a "not yet" into a developer story, and this blog tracks all of them on the Tracker. A developer page or documentation site from Block. Published pricing, whether per minute, per session or as an on-device licence. API keys, SDKs or a console appearing for anyone beyond safety researchers. Or released weights, which would make "API" the wrong question entirely — developers would simply run it themselves.
Until one of those appears, treat third-party "Dessa API" listings, wrappers and resellers as what they are: unaffiliated, and unable to sell you access to a model that has no public access at all.
The bottom line for developers
There is no Dessa AI API, no announced date for one, and no legitimate way to build on Dessa today. Join the free waitlist at block.xyz/dessa if you want Block's own updates, build voice features on platforms that exist now if you have a product to ship, and let the safety-research phase do its work. When Block announces developer access — if it does — we will publish a getting-started guide the same day.
Sources: Block's official Dessa page (block.xyz/dessa) for the research-preview status, local inference and access order; DessaAI Daily Search Console data, October 31, 2026, showing developer API queries; The Verge and TechCrunch, October 2026.
Frequently asked questions
Is there a Dessa AI API?
No. Block has not announced a Dessa API. There is no documentation, API key, developer console or pricing. Dessa is a research preview with a waitlist and early access for safety researchers first.
How much does the Dessa API cost per minute?
There is no Dessa API and no pricing of any kind has been announced for Dessa. Anyone quoting a per-minute Dessa API price is not affiliated with Block.
What can developers use instead of a Dessa API today?
OpenAI's realtime voice offering and Google's Gemini Live provide developer access today, and Kyutai's Moshi offers open weights for a full-duplex system you can run and inspect yourself.
🚀 Get Dessa launch alerts
Dessa is still a research preview. Join the free DessaAI Daily list and we will email you the day access opens — plus the week's biggest Dessa stories, once a week. No spam, no hype.
Form not loading? Subscribe on the signup page →