The first fully general computer action model
We pretrained a model on our 11-million-hour video dataset. Our model can explore complex websites, complete multi-action CAD modeling sequences, and drive a car in the real world, all at 30 FPS.
The Heap: how to rack your own storage
Documenting how we built a 30 petabyte data storage cluster in downtown San Francisco for under half a million dollars.
An open base model for conversational speech
We're releasing an 8.5B parameter autoregressive model for interactive, full-duplex speech generation.