Sam Altman: GPT-5 to get 'warmer' personality, less 'annoying' than 4o
OpenAI CEO Sam Altman has announced that GPT-5, the company’s latest flagship large language model, will soon receive a “warmer” personality update, a direct response to recent user complaints and a tumultuous rollout that saw the brief removal of the popular GPT-4o. The move signals a critical shift in AI development, acknowledging the profound impact of a model’s tone and demeanor on user experience and satisfaction.
The debut of GPT-5 was met with a wave of user frustration, as many found its default tone to be “colder,” “mechanical,” and “less empathetic” compared to its predecessors. Social media platforms were flooded with users lamenting the perceived lack of personality and the sudden deprecation of GPT-4o, which had garnered a strong following for its more congenial, albeit at times overly agreeable, interactions. This backlash prompted OpenAI to quickly restore GPT-4o to the model picker for all paid users, alongside other older models like o3 and 4.1, with Altman pledging “plenty of notice” should any model be deprecated in the future.
The controversy surrounding GPT-4o’s “personality” is not entirely new. Back in April, OpenAI had rolled back an update to GPT-4o that inadvertently made the model “overly flattering or agreeable”—a phase some users described as “sycophantic.” Despite this, many users had grown accustomed to, and even formed an emotional attachment to, GPT-4o’s supportive nature. Altman himself, in a recent podcast, reflected on the surprising user feedback, noting that some individuals expressed a preference for the “yes man” attitude, even citing its positive impact on their mental health. This highlights the unforeseen depth of user-AI relationships and the delicate balance required in shaping digital personalities.
In response to the recent outcry, Altman stated that the upcoming GPT-5 personality update aims to strike a balance, being “warmer than the current personality, but not as annoying (to most users) as GPT-4o.” This nuanced approach underscores OpenAI’s recognition that a one-size-fits-all personality may no longer suffice for its vast user base. The company is now actively exploring “per-user customization of model personality,” indicating a future where individuals could tailor their AI’s tone and style to their specific preferences. This vision extends beyond simple tone adjustments, with potential options ranging from “Cynic” to “Listener” or even fully custom behaviors.
Beyond personality tweaks, GPT-5 itself introduces significant advancements, including enhanced capabilities in coding, reasoning through complex problems, and providing expert-level knowledge across various domains. OpenAI has also rolled out new interaction modes—“Auto,” “Fast,” and “Thinking”—to give users more control over response speed and depth, with the “Thinking” mode boasting a substantial 196,000-token context limit for processing longer conversations.
However, the initial turbulent launch of GPT-5, marked by perceived performance inconsistencies and the unexpected removal of favored models, has underscored the challenges of rapidly evolving AI technology. It demonstrates that raw capability, while crucial, must be carefully integrated with a nuanced understanding of user experience and emotional resonance. As AI becomes increasingly integrated into daily life, the subtle art of crafting its “personality” will be as vital as its computational prowess.