Block launches Dessa, a full‑duplex voice model
Block unveiled Dessa, a speech‑to‑speech AI that runs locally on consumer‑grade hardware, though it isn’t publicly available yet.
Image: Block
Block released a new voice model called Dessa, a full‑duplex speech‑to‑speech system that runs on consumer‑grade hardware such as a MacBook M4 Max or better[1].
Dessa is trained entirely on human speech and runs on a single NVIDIA GPU, enabling it to generate spoken responses while listening to the user in real time[1]. The company says the model can be deployed locally, which sidesteps latency and privacy concerns associated with cloud‑based speech services.
"We believe voice is the most natural way to interact with intelligence," Block said[1].
Despite the technical promise, Block has not opened public access to Dessa yet, positioning the model as a potential internal tool or future offering[1]. The move aligns with Block’s broader focus on providing finance solutions to small businesses[1].
Questions and answers
What does full‑duplex mean for a voice model?
Full‑duplex allows the system to listen and speak simultaneously, creating a more natural back‑and‑forth conversation.
Can developers use Dessa right now?
No, the model is not publicly available at this time.
Companies & investors in this story
Sources
- 1.Block’s new voice model sounds eerily human — TechCrunch AI, Oct 24, 2026
