According to The Verge, OpenAI reportedly plans to release its next frontier AI model, codenamed Orion, by December of this year.
Unlike the previous OpenAI’s two models, GPT-4o and o1, Orion won’t be released widely through ChatGPT; instead, the company plans to grant the companies access to work with the model closely and build their own products and features.
A source familiar with the plan has revealed that the engineers inside Microsoft and OpenAI’s main partner in deploying AI models are preparing to host the Orion on Azure soon in November. Although Orion is a successor to GPT-4, we have yet to determine whether the company would call it GPT-5 publicly.
Although an OpenAI spokesperson has said the reports are inaccurate, they have refused to further elaborate on the rumors.
An OpenAI executive teased the next flagship prowess on the X account, revealing that the Orion is 100 times more powerful than GPT-4, and it is separate from the o1 reasoning model.
The company’s ultimate goal is to combine LLMs and create a powerful model that would be called artificial general intelligence, or AGI.
Previous reports revealed that OpenAI was using o1, code-named Strawberry, to provide artificial data to train Orion. In September, the OpenAI researchers disclosed they successfully trained the model.
It aligns with a cryptic post on X by the OpenAI CEO Sam Altman, where he said, “excited for the winter constellations to rise soon.” When you ask the ChatGPT o1-preview what Altman’s post is hiding, it will hint at the word Orion, the most visible winter constellation from November to February.
Reports say this news has come at a great time for OpenAI, which recently closed a historic $6.6 billion funding round, requiring the company to rebuild itself.
