I'm surprised they didn't benchmark it against Pixtral. They test it against a bunch of different Multimodal LLMs, so why not their own? I don't really see th

I'm surprised they didn't benchmark it against Pixtral. They test it against a bunch of different Multimodal LLMs, so why not their own? I don't really see the purpose of the OCR form factor, when you have multimodal LLMs. Unless it's significantly cheaper.