ChatGPT handling complex OpenAPI schemas

I have an OpenAPI schema where I have more than 50 request types but the response model is same for all of them them. I wonder if having a single endpoint that uses discriminator feature is better than having 50 different endpoints. Does anybody have experience?