Last time I used GPT-5 mini it seems much slower than the primary GPT model API when we used it for an AI chat agent. Particularly around streaming the response

Last time I used GPT-5 mini it seems much slower than the primary GPT model API when we used it for an AI chat agent. Particularly around streaming the responses. But everything I've read implies it's supposed to be faster.