Image recognition accuracy

I think I read somewhere that Chipp had turned on OpenAI's image recognition as default. However, is that right? When configuring an app, the switch to turn it on is still described as using Gemini. I just uploaded a Maths textbook image of several graphs from my phone to a Chipp app and also natively to Chat-GPT. Chat-GPT read it correctly, but the Chipp equivalent got some of it wrong. What’s the best recognition we can get and how?