Hey everyone, We are looking into this issue with our team. We will provide an update very shortly.
Yes, I first noticed it on the weekend from 22th august 2026 as well, maybe went unnoticed by myself the previous two days, but the date range seems to match urs reported beginning to take over the dumb 5.5 mini instant model while 5.6 Sol Thinking was selected which is faking the real reported model to users generated messages, as it still showed “5.6 Sol used” i the hover info on mesages generated by mini model, which is a lie
can you elaborate or provide example? because I only ever noticed it in the “Chat” mode of ChatGPT never in codex or work. Howto reproduce so I could check if affected? I mean, with non distinctive responses its never 100% compareable anyway, but the resolved_model_slug`: gpt-5.5-mini was very ovivously wrong
I can’t just point out that “it says here that it’s a 5.5 mini” or something like that. OpenAI performs these rotations silently on the server side without returning such information to the client. But if you run the same task or benchmark on Codex within your subscription limits and on Codex via the API, where you pay per token, you’ll notice a difference—especially on the benchmark—which confirms that the model has been swapped out.
Hey everyone, Our team has worked on improving this behaviour. Access to certain models or features may be temporarily limited based on account activity, even when a paid subscription is active. We are unable to provide additional details about these checks.
Please review the Terms of Use and the model and feature access troubleshooting guidance. Access is automatically reassessed and can return to normal once the activity affecting availability stops. Hope this helps. We will now close this thread as we have worked on remediation and this is not something that will completely stop as stated reviewing the Terms of Use and the model and feature access troubleshooting guidance should help but if you still face prolonged issue work with our support team (support@openai.com) individually via support cases. Thank you!