I recently tried to run batched API calls using gpt-5-mini for 5,000 requests. Each request has around 10k input and 10k output tokens when I’ve tested using the regular API, but for some reason OpenAI registered 15,000 requests and my bill was over $200. I calculated that this batched processing would cost no more than $10. On top of this, 200/5000 requests were actually completed. I thought we only get charged for completed calls and not failed calls? Has anyone ran into any issues with gpt-5-mini and batch calling? Or has anyone successfully reached OpenAI support? I got a response using their chatbot that they will get back within 2-3 days, is this reliable?