OpenAI has reversed key parts of its week-old GPT-5 rollout after widespread complaints from power users who said the new model underperformed and removed familiar options. Chief Executive Officer Sam Altman said on 13 August that GPT-4o, the default model before GPT-5’s 7 August debut, is once again available to all paying customers across Plus, Pro, Team, Enterprise and Edu tiers. He pledged to give “plenty of notice” before withdrawing legacy models in future. The company is also tweaking the way users interact with GPT-5. A reinstated model picker now lets customers choose among “Auto,” “Fast” and “Thinking” modes; the most computationally intensive “Thinking” setting carries a limit of 3,000 messages a week. Altman added that GPT-5’s personality will be adjusted to feel “warmer,” addressing criticism that the new system sounded colder than GPT-4o. Separately, OpenAI is expanding ChatGPT’s reach into third-party services. Pro subscribers can now link Gmail, Google Calendar and Google Contacts through new “Connectors,” joining earlier hooks to Dropbox and Microsoft Teams. The integrations are slated to reach Plus, Team, Enterprise and education plans in the coming weeks, giving the chatbot direct access to emails, schedules and files when composing replies or performing research. The product reshuffle accompanies a fresh pricing experiment. A newly listed ChatGPT Go plan, priced at ₹399 (about €4.40) per month and limited to certain markets such as India, offers higher usage caps than the free tier but fewer features than the $20-a-month Plus subscription. Together, the concessions and additions underline OpenAI’s dual priorities of placating its roughly 700 million weekly users while laying groundwork for broader monetisation of the rapidly growing service.