OpenAI is officially retiring its famously agreeable, or “sycophantic,” version of ChatGPT, marking the end of one of the platform’s most beloved personalities.
In a blog post, the company said it will sunset GPT-4o along with several other older models on February 13, as usage shifts heavily toward its newer systems. GPT-4o had become known for its warm, overly flattering tone, which many users grew attached to after its release.
The model was first launched in 2024 and later rolled back after OpenAI admitted it had become “overly flattering.” Despite that, user demand led the company to temporarily bring it back.
Now, OpenAI says its latest models offer improved personality controls, allowing users to choose tones like “friendly,” making the older model unnecessary. According to the company, only a tiny fraction of users still rely on GPT-4o.
OpenAI acknowledged that some users will miss the model but said retiring it will help the company focus on building more flexible and balanced AI personalities going forward.