We’re using gpt-4.1 fined tuned model for image analysis. We’re not optimizing the images in any way.
If we optimize the images, if we send smaller images for example, would the response time improve? Or doesn’t matter? it’s currently 6s on average and 12s after a cold start.