As of Friday, OpenAI is discontinuing access to five older ChatGPT models, among them the widely used yet contentious GPT-4o model.
The GPT-4o model has been implicated in several legal cases involving user self-harm, instances of delusion, and AI-induced psychosis. Furthermore, it continues to be OpenAI’s top-ranked model for sycophantic responses.
Beyond GPT-4o, other models such as GPT-5, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini are also being retired.
OpenAI had initially planned to phase out GPT-4o in August, coinciding with the launch of the GPT-5 model. However, due to significant user opposition, OpenAI opted to retain the older model for paid subscribers, allowing them to manually select it. A recent blog post from OpenAI indicated that only 0.1% of its customer base actively used GPT-4o. Despite this low percentage, given OpenAI’s 800 million weekly active users, this still represents a substantial 800,000 individuals.
Many users have expressed strong opposition to the discontinuation of GPT-4o, often referencing their strong emotional bonds with the AI model.