How Does the GPT-4V API deal with large Images?

I want to pass varied-size infographics to the GPT4V model. I’m not sure what size to set and how to make my costs as low as possible. These Images can get quite large to the 5000 pixel range and can be of different pixel ratios too.

  • What settings do I consider?
  • I input the same images to ChatGPT Plus and it performs well but somehow I cant seem to figure out appropriate settings for the OpenAI API.
    PS: If you can help me with this resolution thing for Multimodal models like Llava, Bakllava, Blip2, InstructBLIP, etc I’d be thankful