OpenAI plans to launch the Orion AI model in December.
Altman has hinted that the launch of Orion is approaching.
OpenAI plans to launch its next model, called Orion, before December. Unlike the previous releases of its GPT-4o and o1 models, Orion will not be widely distributed from the outset through ChatGPT. Instead, the company intends to initially grant access to companies it closely collaborates with, so they can develop their own products and functionalities.
Sources indicate that engineers at Microsoft, OpenAI's main partner for implementing artificial intelligence models, are preparing to host Orion on Azure starting in November. While Orion is expected to be the successor to GPT-4, it is unclear if OpenAI will choose to publicly name it GPT-5. It’s important to note that the release plan could be subject to changes and delays. Both OpenAI and Microsoft refrained from commenting on this situation.
An OpenAI executive has hinted that Orion could be up to 100 times more powerful than GPT-4 and that it is different from the reasoning model o1 launched in September. The company aims to merge its large-scale language models over time, with the goal of creating an even more powerful model that could be considered artificial general intelligence (AGI).
It was previously reported that OpenAI was using o1, code-named Strawberry, to generate synthetic data to train Orion. In September, OpenAI researchers held an informal meeting to commemorate the completion of training for this new model, which aligns with an enigmatic message on X from the company's CEO, Sam Altman, who expressed his excitement about the arrival of the winter constellations.
The model's launch comes at a decisive moment for OpenAI, which recently completed a historic funding round of $6.6 billion, forcing the company to restructure as a for-profit entity. Additionally, the company is facing considerable turnover, with the departure of CTO Mira Murati, head of research Bob McGrew, and Barrett Zoph, VP of post-training.
