OpenAI has swiftly responded to user criticism after replacing GPT-4o with its newer GPT-5 model as the default in ChatGPT. Many subscribers complained that GPT-5 felt “cold,” “mechanical,” and “less empathetic” compared to its predecessor, prompting threats of subscription cancellations. In response, the company reinstated GPT-4o to the model picker for all paid users, while also promising updates to GPT-5’s tone and personality.
Chief Executive Officer Sam Altman confirmed that GPT-5 will soon receive a warmer personality update that avoids what some users found “annoying” about GPT-4o. He acknowledged that the rollout highlighted the need for greater per-user customization, allowing individuals to adjust the AI’s style and tone to their liking.
OpenAI has also introduced three modes for GPT-5: Auto for general use, Fast for speed, and Thinking for deep reasoning, which offers an expanded 196,000-token context window and a weekly limit of 3,000 messages for paid users. Once the limit is reached, access switches to the lighter GPT-5 Thinking mini.
Alongside restoring GPT-4o, OpenAI has made older models like GPT-4.1, GPT-3.5, and GPT-5 Thinking mini accessible through a “Show additional models” setting, while GPT-4.5 remains exclusive to Pro subscribers due to its GPU demands. Altman noted that the move reflects how strongly users connect with certain AI personalities, a factor the company had underestimated. This emotional connection, he suggested, will guide future development priorities.
OpenAI plans to expand customization options, potentially allowing users to choose from preset personalities such as “Cynic,” “Listener,” “Robot,” or “Nerd,” or even create fully tailored styles and behaviours.
Altman said these changes mark a broader shift in AI development, where emotional resonance and user experience are as critical as raw technical capability. “We’re not always going to get everything on try #1,” he remarked, “but I’m proud of how quickly the team can iterate.”