OpenAI confirms GPT-4o 'retirement', citing low usage and focus on new models

ChatGPT

• OpenAI to retire GPT-4o, GPT-4.1, and o4-mini from ChatGPT soon
• Only 0.1 percent of users chose GPT-4o daily; most now use GPT-5.2
• API access to retired models remains unchanged for developers

Did our AI summary help?

OpenAI said on Thursday that it plans to retire GPT-4o from ChatGPT, alongside GPT-4.1, GPT-4.1 mini and OpenAI o4-mini. The decision comes less than a year after GPT-4o was launched in May 2024 and gained popularity among a subset of paid users for its warm, natural conversational style.

The company acknowledged that the move would disappoint some users but said it was necessary to focus resources on the models most people are actively using today.

A model users fought to keep

GPT-4o became a point of tension last year when OpenAI briefly removed access to it following the launch of GPT-5. That decision sparked backlash from users who preferred GPT-4o’s tone and interaction style. OpenAI quickly reversed course, restoring access for paid users.

At the time, CEO Sam Altman said the company would give “plenty of notice” before retiring GPT-4o. With this announcement, OpenAI appears to be following through on that promise.

Why OpenAI says it’s making the change

According to OpenAI, only around 0.1 percent of users choose GPT-4o as their daily model, while the “vast majority” of users now rely on GPT-5.2. The company said recent improvements to model personality, customisation and creative ideation have made it confident that newer models can meet most users’ needs.

“Retiring models is never easy,” OpenAI said in its blog post, adding that streamlining its lineup allows it to improve the experience for most users more quickly.

GPT-4o is not the only model being phased out. OpenAI also confirmed that GPT-5 Instant and GPT-5 Thinking will be removed from ChatGPT, part of a broader effort to simplify the chatbot’s model selection and reduce fragmentation.

The company clarified that these changes apply only to ChatGPT. There are no updates to OpenAI’s API offerings, meaning developers can continue using existing models through the API for now.