OpenAI is set to discontinue the original API snapshot of GPT-4o, released on May 13, 2024, by October 23. The company officially retired the model on February 13, 2026, citing that only 0.1% of users were still selecting it over newer versions like GPT-5.2.
The decision sparked a long-standing campaign by users to preserve the model. Petitions on Change.org gathered over 20,000 signatures, with supporters arguing that the model provided unique emotional support and conversational qualities that newer iterations lacked. In February 2026, protesters held a vigil at OpenAI's San Francisco headquarters.
Advocacy groups, including an Austrian digital rights organization, have formally requested that OpenAI release the model's weights under an Apache 2.0 license. They argue that internal safety assessments from August 2024 classified the model as low-risk, justifying its public release for research purposes.
The development of GPT-4o involved key figures who have since moved on. Alexis Conneau, who envisioned the model's emotional voice capabilities, left OpenAI in 2024 and joined Meta following the acquisition of his startup, WaveForms AI. Other core team members have transitioned to new ventures or internal roles within OpenAI.
In contrast to OpenAI's approach, competitor Anthropic maintains a policy of preserving weights for all publicly released models. Anthropic also conducts 'retirement interviews' with models before they are decommissioned, documenting their responses to their own development and usage history.
