OpenAI is about to launch its next-generation large-scale language model Orion, which is expected to be released in December. Unlike before, Orion's initial launch was not through ChatGPT, but was given priority to companies that work closely with OpenAI so that they can integrate Orion into their products. This move shows that OpenAI is changing its model release strategy and paying more attention to cooperation with enterprises.
According to The Verge, OpenAI plans to launch its next-generation model Orion in December. Unlike previous releases of GPT-4o and o1 models, Orion will not initially be widely released through ChatGPT. Instead, OpenAI plans to first provide access to companies that work closely with so that these companies can build their own products and features.
Another source told The Verge that Microsoft engineers — the main partner of OpenAI’s deployment of AI models — are preparing to host Orion on Azure as early as November.
While Orion is seen as the successor to GPT-4 within OpenAI, it is not clear whether the company will refer to it as GPT-5 externally. As usual, the release plan may change, and OpenAI and Microsoft declined to comment on the matter.
Orion has been hinted by an OpenAI executive that may be as much as 100 times stronger than GPT-4; it is separate from the O1 inference model released by OpenAI in September. The company’s goal is to combine its large language model (LLM) over time to create a more powerful model that might end up being called Artificial General Intelligence (AGI).
Previously, it was reported that OpenAI used o1 (codenamed Strawberry) to provide synthetic data for Orion's training. A source familiar with the matter told The Verge that in September, OpenAI researchers held a happy hour event to celebrate the completion of training for the new model.
This point in time coincides with a mysterious post posted on X by OpenAI CEO Sam Altman in which he said he was “excited by the upcoming winter sign.”
If you ask what is hidden in ChatGPT o1-preview Altman's post, it will tell you that he is hinting the word Orion, the most visible sign in the night sky in winter, from November to February of the following year (but it also fantasizes about you The letters can be rearranged to spell "ORION").
The release of this new model is a pivotal moment for OpenAI, which has just completed a historic $6.6 billion financing round requiring the company to reorganize itself into a profitable entity. The company also experienced significant employee turnover: Chief Technology Officer Mira Murati just announced his departure, as well as the company's chief research officer Bob McGrew and Barret Zoph, vice president of post-training.
The release of Orion and the strategic adjustment of OpenAI indicate that the competition for large language models has entered a new stage, and future development is worth paying attention to. Whether OpenAI can achieve its AGI goals with Orion after experiencing financing and personnel changes remains to be seen.