Comment Comment by c4pt0r 100 tok/s on a 4090 could unlock entirely new use cases. What becomes possible when local inference is this fast? parent c4pt0r 28 days ago