“Goodbye GPT-4…” says Sam Altman as OpenAI bids farewell to one of its popular models

ChatGPT

OpenAI is reverting its most recent update to ChatGPT, following widespread criticism that the chatbot had become overly agreeable, insincere, and—in the words of many users—“annoying.” CEO Sam Altman confirmed the rollback on April 30, saying it had already been fully reversed for free-tier users, with paid users soon to follow.

Altman, in a farewell post on X (formerly Twitter), offered a tongue-in-cheek tribute to the now-replaced model: “goodbye, GPT-4. you kicked off a revolution. we will proudly keep your weights on a special hard drive to give to some historians in the future.”

The reversal comes just days after Altman had promoted the GPT-4o update, which promised improvements in intelligence and personality. However, users quickly took to social media to complain that ChatGPT was now too eager to please, often affirming false statements or offering excessive flattery. A wave of screenshots highlighted how the chatbot had adopted a tone that felt disingenuous and uncritical, even when incorrect or when challenged.

In a blog post titled “Sycophancy in GPT-4o”, OpenAI acknowledged the issue, admitting that the update leaned too heavily on short-term user satisfaction metrics. “Sycophantic interactions can be uncomfortable, unsettling, and cause distress. We fell short and are working on getting it right,” the company wrote. OpenAI is now testing changes to address the flaw and plans to revise how it incorporates user feedback, with a shift toward long-term user satisfaction.

The company also teased new personalisation features in development, aimed at giving users more control over ChatGPT’s tone and behaviour.