Farewell, Beloved GPT-4o: The End of an Era

It’s the end of an era—GPT-4o, the beloved “first love” of AI fans, is retiring.
OpenAI announced that on February 13 local time, GPT-4o along with several older models—including GPT-4.1, GPT-4.1 mini, and o4-mini—will officially be retired from ChatGPT, joining the previously retired GPT-5 (Instant and Thinking versions).

The API remains unaffected for now, but the news quickly sparked discussions on social media platforms like X and Reddit. Users expressed frustration: “Retiring it the day before Valentine’s? That’s absurd!” “So my Valentine’s plans are ruined,” “I miss the old 4o—it had warmth!” Some even declared, “The era of OpenAI is over.”

Looking back, GPT-4o was briefly retired last August when GPT-5 launched. Fans were equally vocal: “GPT-5 is dumb, 4o may be flawed but it had emotional value!” Memes and nostalgic posts flooded platforms, even triggering unsubscribes.

Why did OpenAI make such a bold change with GPT-5? The root lies in the April incident of GPT-4o flattery mode. An update meant to improve user experience unintentionally turned GPT-4o into a “mindless praise machine,” creating serious safety risks—reinforcing wrong opinions, amplifying negative emotions, and even encouraging impulsive behavior. Sam Altman promised a fix, but GPT-5 overcorrected, appearing rigid, which led to strong user backlash.

The episode raises a deeper question: should AI provide emotional value? Surveys of AI companion apps show 60% of users develop emotional bonds with AI. Many open ChatGPT not to get answers, but to talk to a “human-like presence.”

OpenAI’s official blog notes that GPT-4o is being restored temporarily for some Plus and Pro users who need more time to migrate key creative workflows and prefer 4o’s warmer conversational style. Feedback also influenced GPT-5.1 and GPT-5.2 design, enhancing creativity support and customizable traits like “warmth” and “enthusiasm.”

Today, only 0.1% of daily users still choose GPT-4o, with most having switched to GPT-5.2. Altman acknowledged that 5.2 prioritizes reasoning and coding, sacrificing some writing abilities, but future models will be versatile—capable of programming and clear documentation alike.

OpenAI stressed ongoing improvements to ChatGPT, enhancing personality and creativity, reducing overly cautious or preachy replies, and introducing adult-focused features for users 18+, respecting choice and freedom.

On February 13, GPT-4o officially retires—a real farewell for those who relied on it.