Reported by @short mountain
Bug Report: GPT-5 Thinking & o3 not triggering reasoning in any session
`Steps to Reproduce`
- Open a new chat session on ChatGPT (Web or iOS).
- Manually select either “gpt-5-thinking” or “o3” model.
- Input a basic reasoning prompt, e.g.:
“The sum of two numbers is 20. One number is 4 more than twice the other. What are the two numbers?” - Observe the model’s behavior.
`Expected Result`
The model should initiate chain-of-thought reasoning: show step-by-step calculations or logic, along with the animated reasoning indicator. The output should reflect structured problem solving, as previously expected from these models in reasoning mode.
`Actual Result`
No reasoning is triggered. No animation appears. The model often gives shallow or irrelevant answers. In some cases, it outputs broken placeholder logic (e.g. “b = ? - a”), and incorrectly identifies itself (e.g. o3 says it’s GPT-4o-mini). This happens in 100% of sessions since Aug 29.
`Environment`
- Web: Windows 10, Chrome v116+ - iOS: ChatGPT App v1.2025.229 - ChatGPT Pro account, manually selecting reasoning models (gpt-5-thinking & o3) - Issue present across all prompts, windows, and devices