OpenAI has given ChatGPT users more control over AI model selection. Now, you can manually switch GPT-5 between “Auto”, “Fast”, and “Thinking” modes. According to Sam Altman, most people will stay on “Auto”, but the new options allow for more flexible model selection for specific tasks.
The option to choose the GPT-4o model has returned for all paid users, and the company has promised to notify about future changes in advance. A “Show additional models” button has appeared in the settings, providing access to older versions, including o3, 4.1, and GPT-5 Thinking mini. The GPT-4.5 model remains available only to Pro subscribers due to its high maintenance cost.
Users can send up to 3,000 requests per week in “GPT-5 Thinking” mode, after which the system automatically switches them to a less powerful version. OpenAI notes that this is a temporary limit increase and warns of possible changes in the future depending on the load.
The ChatGPT interface also received a minor update — if you hover over the “Regenerate” button, you can see which model responded to the request. This helps users better understand how model selection works during interaction with the service.
