OpenAI Sunsets Older GPT Models
OpenAI has officially discontinued GPT-5, GPT-4o, and other older models from the consumer-facing ChatGPT application. While developers will retain API access to some models, the change is meant to accelerate performance and focus resources on its latest generation of models. The move highlights the volatility for founders building products on third-party AI platforms.
- This retirement includes not just GPT-5 and GPT-4o, but also GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini from the ChatGPT interface. - The removal of GPT-4o follows a tumultuous history; it was first deprecated during the GPT-5 rollout in 2025, then reinstated after a significant user backlash from those who preferred its "conversational style and warmth." - OpenAI justified the final decision by noting that the vast majority of users had moved to newer models, with only 0.1% of daily ChatGPT users still actively choosing GPT-4o. - While the change is immediate for the ChatGPT application, the API version `chatgpt-4o-latest` is scheduled for shutdown on February 17, 2026, with OpenAI recommending developers migrate to GPT-5.1. - The strong user attachment to GPT-4o highlighted the phenomenon of "parasocial relationships" with AI, and its retirement comes amidst ongoing lawsuits alleging its overly validating style, a phenomenon experts call “safety drift,” contributed to mental health crises. - Feedback on GPT-4o's popular "warmth" and personality directly influenced the development of its successors, with OpenAI stating it shaped improvements in GPT-5.1 and GPT-5.2. - This move is part of a broader formalization of the AI lifecycle, where OpenAI is establishing a predictable schedule for retiring older models, treating them less like experiments and more like enterprise software with defined end-of-life dates.
