‘Suddenly deprecating old models’ users depended on a ‘mistake,’ admits OpenAI’s Altman

OpenAI brought back GPT-4o following a weekend of user complaints – mainly about the removal of model choices – after the rollout GPT-5.

GPT-5 was launched last week. OpenAI made much of the improvements, and claimed that the update would reduce hallucinations. GPT-5 is not a single model but a collection. Prompts are routed to the models based on factors such as intent and complexity.

Simple? Not quite. OpenAI removed the option to choose older models such as GPT-4o () because GPT-5 can handle the decision for you. Users were furious, as they had become dependent on older models for their workflows. Customers have grown accustomed to the strengths and weaknesses of each model.

Switching from one model to another “to rule them all” led to a cacophony complaints and, surprising for a tech firm, a U turn by OpenAI.

Open AI’s CEO Sam Altman confirmed to a user who asked if 4o would be returning: “it’s back! go to settings and pick ‘show legacy models’.”

Altman acknowledged that users were disgruntled and the subsequent furor.

“It feels different and stronger than the kinds of attachment people have had to previous kinds of technology (and so suddenly deprecating old models that users depended on in their workflows was a mistake).”

“Mistake” would be one way to put it.

A user asked: “Will I be marked as weird if I simply have Legacy Models on? Or only if I use 4o a great deal? OpenAI pulls the plug on ChatGPT smartbot that praised a user for quitting psychiatric medication