[energy123]

It's possible they have a software layer that does that. But I was assuming they don't, because the open source multimodal models don't.