OpenAI is re-tooling its flagship chatbot only a week after releasing GPT-5, acknowledging user backlash over the abrupt withdrawal of earlier models and confusion about the new line-up. In an interview published 14 August, Nick Turley, the company’s vice-president and head of ChatGPT, said the decision to drop GPT-4o was a “miss” and confirmed the model will return for Plus subscribers while OpenAI works on a predictable deprecation schedule. Power users will also regain the option to pick from a full list of models instead of relying solely on GPT-5’s automatic selection. Turley disclosed that ChatGPT now serves about 700 million people each week and counts 20 million paying subscribers and roughly 5 million business customers. Yet only 7 % of paying users and 1 % of free users had been invoking so-called “reasoning” modes before GPT-5’s launch, a pattern that contributed to frustration when the faster but less capable GPT-4o vanished. The company is therefore revising default behaviour: responses will be shorter, more direct and less focused on sustained conversation, with over-use prompts and a new ‘personality’ picker to let users adjust tone without relying on an ever-growing list of model names. Looking ahead, OpenAI plans to give ChatGPT “deeper context and wider actions,” allowing the assistant to observe more of a user’s workflow and trigger external tools. Turley said the team is also evaluating advertising and affiliate commerce as possible revenue streams but intends to preserve the subscription model that generated rapid growth. He emphasised that future changes will be announced earlier and that older models will remain available for a defined period to avoid a repeat of this month’s disruption.