Does this matter if they do it on release or at a later date?

@RemNi By integration, i meant something that is quite fundamental, possibly during pretraining and architecture design. So, even if it is not released on a later date, we will likely know if they were integrated.

A GPT4o level integration of video generation and language suffices.

@Sss19971997 ah think I understand what you mean. In that case I reckon no, I suspect GPT-5 will be a multimodal autoregressive transformer, and will not integrate a diffusion component

@Sss19971997 maybe update the description to make this a bit more detailed

GPT4o would resolve YES, but I have no idea how they did it.