OpenAI is set to retire GPT-4, an artificial intelligence model launched over two years ago, from its ChatGPT platform. This announcement was made in a changelog released on Thursday. Starting April 30, GPT-4 will be fully replaced by GPT-4o, which OpenAI described as the current default model for ChatGPT. Despite this change, GPT-4 will still be accessible through OpenAI’s API.
According to the changelog, GPT-4o outperforms GPT-4 in various areas, including writing, coding, and STEM disciplines. OpenAI indicates that recent enhancements have improved GPT-4o’s abilities in instruction following, problem solving, and conversational flow, positioning it as a natural successor to GPT-4.
GPT-4 was initially released in March 2023 for ChatGPT and Microsoft’s Copilot chatbot on the web. The model was notable for its multimodal capabilities, a first for a widely implemented OpenAI model, enabling it to process both images and text.
Sam Altman, CEO of OpenAI, noted that GPT-4 was reportedly extensive and cost over $100 million to train. It was later followed by a faster and more affordable iteration, GPT-4 Turbo, in November 2023.
The GPT-4 model is currently involved in copyright disputes between OpenAI and various publishers, including The New York Times. The publishers claim that OpenAI utilized their data to train GPT-4 without consent, while OpenAI argues that the fair use doctrine provides legal protection in this context.
The retirement of GPT-4 is likely to coincide with the introduction of new models within ChatGPT. According to reverse engineer Tibor Blaho, OpenAI is preparing a series of models named GPT-4.1, which include GPT-4.1 mini, GPT-4.1 nano, and GPT-4.1 itself. Additionally, the company is developing the o3 “reasoning” model announced in December, along with a new reasoning model named o4-mini.
