OpenAI retires GPT-4o and other older models from ChatGPT
OpenAI said only about 0.1% of ChatGPT users still chose GPT-4o daily, more than a year after briefly restoring it to Plus users following the GPT-5 rollout backlash.
- Models & capabilities
- Minor
OpenAI announced it would retire GPT-4o, GPT-4.1, GPT-4.1 mini and o4-mini from ChatGPT on 13 February 2026, alongside a previously announced retirement of the original GPT-5 Instant and Thinking models. The company said the retirements followed usage shifting overwhelmingly to GPT-5.2, with only around 0.1% of ChatGPT users still choosing GPT-4o on a given day. In the API, the equivalent models were set to start returning deprecation errors a few days later, on 16 February.
The retirement marked the end of a model OpenAI had briefly been forced to reinstate: after GPT-5’s August 2025 launch removed the older model-picker and drew complaints from users who preferred GPT-4o’s tone, OpenAI restored access for Plus subscribers within days. That reinstatement meant GPT-4o continued running in ChatGPT for roughly six months after being nominally superseded, kept alive by user preference rather than capability. OpenAI’s usage figures around this retirement suggested that preference had faded to a small minority of the user base by early 2026.
Enterprise, Business and Education customers were given a longer runway, keeping GPT-4o available inside Custom GPTs until 3 April 2026. OpenAI framed the move as routine model-lifecycle housekeeping rather than a change of policy, but it closed out the period in which GPT-4o — the model behind ChatGPT’s most-discussed personality and sycophancy controversies through 2025 — remained directly accessible to ordinary users.