Block unveils Dessa, a speech model trained only on human conversation

Block's AI lab introduced Dessa, a full-duplex speech-to-speech model that listens and talks at once. Block says it was trained from scratch on human speech only, with no text or pretrained models, on one Nvidia GB200 NVL72, and runs locally on a MacBook Pro M4 Max or better. Not public: a waitlist, with early access limited to a small group of AI safety researchers. No release date or license is stated.

HEAT

OUTLETS

14

FIRST SEEN

Oct 23, 2026, 15:32 ET

LAST SIGNAL

17h ago