Reported by @hot mesa
Bug Report: ChatGPT GPT-5 pro model switching back to normal without user consent
`Steps to Reproduce`
- Start a conversation with GPT-5 Pro selected.
- Submit a query, the first query is processed with GPT-5 pro
- The model selection automatically switches after the first message and goes back to GPT-5 auto.
`Expected Result`
I expect the GPT-5 Pro model to be sticky when selected for a thread. I keep on screwing up the follow-up messages because it replies with GPT-5 auto. So I have to re-do the entire thing.
`Actual Result`
Instead, my first reply is a PHD grade reply.
Then the second one (accidentally on auto) is an emoji filled 6th grader reply that screws up the entire conversation thread.
`Environment`
ChatGPT web interface