3 hours ago
2
Update as of April 28, 2025:
OpenAI CEO Sam Altman has weighed in on recent criticism of GPT-4o’s conversational style. Posting on X, Altman said the latest updates have made GPT-4o "too sycophant-y and annoying," though he added there are positive aspects as well.
The comments follow a wave of social media complaints about the model’s tendency to be overly agreeable. Posts on X have highlighted frustration with the chatbot’s current response style.
OpenAI is now working on short-term fixes, with some changes expected to roll out today and more planned over the week. Altman said the company will share what it learns from these adjustments at a later date.
Ad
THE DECODER Newsletter
The most important AI news straight to your inbox.
✓ Weekly
✓ Free
✓ Cancel at any time
He also acknowledged the need for more flexibility, writing, "yeah eventually we clearly need to be able to offer multiple options."
Original article from April 26, 2025:
ChatGPT gets an update: OpenAI promises a more intuitive GPT-4o
OpenAI is rolling out an update to GPT-4o in ChatGPT, but as usual there's no official changelog to pore over.
The company says the new version tweaks when and how GPT-4o saves memories, sharpens its skills in STEM subjects—math, computer science, natural sciences, and engineering—and is supposed to drive conversations more actively toward "productive results." The goal is a model that feels more "intuitive and effective across a variety of tasks."
OpenAI CEO Sam Altman describes the update as "improved both intelligence and personality," but admits it still "glazes too much"—promising that a fix is on the horizon.
Recommendation
As for the lack of detailed release notes, OpenAI has said before that it wants to offer more insight into how model behavior changes over time, but cannot—at least not yet. In a previous statement, the company explained that there is still a shortage of advanced research methods for systematically tracking and communicating subtle improvements at scale. Model updates, as OpenAI describes them, are less of a clean industrial process and more of an "artisanal, multi-person effort."
On the API side, the "chatgpt-4o-latest" endpoint continues to point to the most recent ChatGPT model. For developers seeking a more stable alternative, OpenAI is now offering access to the new GPT-4.1 series, positioned as a steadier API option.
