OpenAI removes access to sycophancy-prone GPT-4o model

Beginning Friday, OpenAI will discontinue access to five older ChatGPT models, prominently including the contentious GPT-4o. This particular model has been embroiled in legal disputes related to issues such as user self-harm and erratic behavior, and has been labeled as the company's most sycophantic model. Alongside the GPT-4o, OpenAI will also retire the GPT-5, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini models. The decision to phase out GPT-4o was initially announced in August when the new GPT-5 model was launched. However, due to substantial public outcry, OpenAI had opted to keep GPT-4o available for paying subscribers who wished to continue using it. Despite the model being utilized by only a fraction of users—0.1% of OpenAI’s 800 million weekly active users, equating to about 800,000 individuals—many have voiced strong opposition to its removal. These users have expressed their deep connections with the model, making the decision to retire it particularly controversial.