OpenAI has restored GPT-4o to the ChatGPT model picker for paid users and added a speed selector for GPT-5 with Auto, Fast, and Thinking modes. The changes are documented in the company's ChatGPT release notes, and CEO Sam Altman announced the selector on August 13, 2025 on X.
Key details
- GPT-4o now appears by default in the model picker for paid accounts, per the ChatGPT release notes.
- OpenAI said it will provide advance notice before removing or deprecating models in ChatGPT, according to the same release notes.
- GPT-5 introduces a mode selector with three options - Auto, Fast, and Thinking - confirmed by Sam Altman.
"You can now choose between 'Auto', 'Fast', and 'Thinking' for GPT-5."
Background
OpenAI introduced GPT-4o on May 13, 2024, as a multimodal model for text, images, and audio, enabling more natural voice and vision interactions. See the announcement: Hello GPT-4o.
OpenAI maintains a public ChatGPT release notes page that logs product updates, including model availability and interface controls.
For broader context on GPT-5's rollout, see Search Engine Journal's coverage.
.svg)
.svg)
.svg)