Diffusion based language model open access from Inception labs

No affiliation. Damn this thing is fast. Seems to use some sort of diffusion based mode combined with some sort of sequencing — text still flows in sequentially, just realllly quickly.

Ignore this - url fat fingered