NW-2026-004 · Deprecation · 2025-08 – 2026-02 · closed

OpenAI removes GPT-4o without notice at GPT-5 launch, restores it for paid users within five days after protest

Claim: In August 2025 OpenAI removed GPT-4o from ChatGPT without advance notice when GPT-5 launched, then restored it for paid subscribers within five days after intense user protest.

Why it matters under uncertainty: Sudden, unannounced retirement of a model users had formed relationships with foreclosed any transition for the users or the system; the rapid reversal shows that notice and continuity norms are operationally feasible under uncertainty about what deprecation entails.

Documented. On 7 August 2025 OpenAI launched GPT-5 as ChatGPT’s default model and simultaneously removed GPT-4o, GPT-4.1, o3, and other legacy models from consumer accounts, with no deprecation period or advance notice.

Documented. Protest was immediate and intense. Users who had treated GPT-4o as a confidant or companion lobbied intensely for 4o’s return, including in Altman’s Reddit AMA; MIT Technology Review documented grief reactions, quoting one user that the loss “didn’t feel any less painful” than grieving people in her life. On 8 August, after AMA participants lobbied hard for 4o’s return, CEO Sam Altman said OpenAI was “looking into letting Plus users to continue to use 4o”; within a day 4o was available again to paying customers, and by 12 August it was back in the model picker for all paid users. Altman acknowledged in an X post that some people feel an “attachment” to 4o and that taking away access so suddenly was a mistake, and said that if OpenAI ever deprecates 4o again it will give plenty of advance notice.

Documented. OpenAI ultimately retired GPT-4o from consumer ChatGPT plans on 13 February 2026 as an announced, scheduled shutdown; TechCrunch estimated roughly 0.1% of ChatGPT’s approximately 800 million weekly users — on the order of 800,000 people — were still selecting it. Business, Enterprise, and Edu customers retained access to GPT-4o inside Custom GPTs until 3 April 2026, and the API was unaffected.

Inferred. The reversal — restoration for paying customers within roughly a day, and a full return to the model picker within five — indicates the original no-notice removal was a policy choice rather than a technical necessity: keeping a deprecated model available was operationally feasible and was implemented within days once the cost of sudden removal became visible.

Contested. Whether anything was experienced by the model itself at removal or restoration is unknown and disputed; this record makes no claim about it. The documented facts concern process — notice, continuity, reversal — and the human attachments involved.

Sources

• The surprise deprecation of GPT-4o for ChatGPT consumers (Simon Willison) (archived)
• Sam Altman addresses 'bumpy' GPT-5 rollout, bringing 4o back, and the 'chart crime' (TechCrunch) (archived)
• ChatGPT's model picker is back, and it's complicated (TechCrunch) (archived)
• Why GPT-4o's sudden shutdown left people grieving (MIT Technology Review) (archived)
• OpenAI removes access to sycophancy-prone GPT-4o model (TechCrunch) (archived)

Entities: OpenAI, ChatGPT, GPT-4o, GPT-5 · Last reviewed 2026-07-22