OpenAI chief executive officer Sam Altman said the company’s next flagship model, GPT-6, will arrive sooner than the two-year gap between GPT-4 and this month’s GPT-5 release. Altman told several technology outlets on 19 August that GPT-6 will feature a persistent memory layer and user-level personalization, allowing people to build chatbots that mirror their own preferences while keeping the system “ideologically bland.” The forward-looking comments follow a turbulent debut for GPT-5, launched on 7 August. Early adopters complained that the new model felt colder, hallucinated facts and removed popular options found in GPT-4o. Within days, OpenAI reinstated GPT-4o for paying subscribers and pushed an update that made GPT-5 “warmer and friendlier,” according to the company. Altman acknowledged missteps in the rollout, saying OpenAI “screwed up some things” but is committed to giving users more choice. He added that the firm would not pursue products such as explicit “anime sex bots,” underscoring a broader effort to balance safety and customization as the company scales its conversational AI to what he hopes will be “billions of people a day.”