Dessa: Block's full-duplex speech model trained only on human conversation
Block's in-house AI research lab announced Dessa on Oct 23, a full-duplex speech-to-speech model trained from scratch on nothing but human conversation — no text, no pretrained model underneath. It listens and talks simultaneously, reacts to interruptions, laughter and backchannels, and supposedly runs faster than real time on a MacBook Pro M4 Max or better. Waitlist opened Oct 23; early access goes first to a small group of AI safety researchers.
Why it matters
The headline training claim: Dessa learned language from listening 'the way a child does', using 'hundreds of times fewer words' than text LLMs — a speech-native path that could preserve timing, tone, and non-text signals that transcription pipelines discard.
What you could build with it
An indie hacker could build a phone-like language-tutoring app on Dessa's full-duplex voice model once access widens — a learner interrupts, hesitates, and laughs while the tutor responds naturally in real time — producing conversational voice practice without the awkward turn-taking of text-pipeline voice assistants.
Does it hold up?
Too early to judge — waitlist opened Oct 23 with access limited to safety researchers; demos show natural full-duplex conversation but no independent evaluation exists yet.
Built with Dessa
- Dessa Waitlist: How to Get Access to Block's Dessa Voice AIjuliangoldie.com · Guide to the Oct 23 waitlist: official signup is block.xyz/dessa, early access goes to AI safety researchers with institutional emails.
- Block introduces Dessa, a speech-to-speech AI model trained on conversationsletsdatascience.com · Independent take cautioning that no public benchmark establishes comparative performance and broad availability is unannounced.
- Block unveils Dessa voice AI model trained on conversations — The Vergeua.news (The Verge) · Reports Block's released demo recordings — including one where Dessa mistakenly addressed its interlocutor as 'Ronaldo'.
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73NVIDIA Isaac Lab 3.0 GA + Isaac OS open runtime + Agile ONE SN humanoidMajor Isaac platform expansion at GTC Berlin Oct 21: Isaac Lab 3.0 is generally available as the open-source robot…models · JEV 0.73StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73NVIDIA Nemotron 3.5 Retriever collection: Embed 8B/1B, Rerank 8B, Parse 631MNVIDIA's Oct 21 collection drop for agentic retrieval: Nemotron-3.5-Embed in 8B BF16, 1B BF16 and 1B NVFP4…models · JEV 0.72Kimi K3.1 Pro: Moonshot's 2.8T open-weight flagship officially launchesMoonshot AI officially launched Kimi K3.1 Pro on Oct 23, a mixture-of-experts model with 2.8 trillion total / 104…models · JEV 0.72
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.