4 ms·The biggest improvements are coming from the diffusion models. Image, video, and voice models.by cynicalpeace 2y agoThe biggest improvements are coming from the diffusion models. Image, video, and voice models.