OpenAI is set to introduce Orion, its latest model, by December, according to The Verge. Unlike previous releases, Orion won’t be immediately available to all users through ChatGPT. Instead, OpenAI plans to give priority access to its close business partners, who will use Orion to build their own tools and features.
OpenAI might release Orion by December
OpenAI is aiming for a more controlled rollout, allowing for better integration and customization by trusted partners before making it widely available to the general public. A source also revealed to The Verge that Microsoft, a key partner, is preparing to host Orion on Azure as early as November. This collaboration highlights the ongoing partnership between OpenAI and Microsoft, which has been a major driving force behind the deployment of AI models at scale.
Orion is viewed internally at OpenAI as the successor to GPT-4, though it’s uncertain if the model will officially be named GPT-5. The model has been described as potentially up to 100 times more powerful than GPT-4, indicating a significant leap in capabilities and performance. OpenAI aims to eventually merge its models into a more advanced system, potentially leading toward artificial general intelligence (AGI), which has been a long-term goal for the company. AGI represents a type of AI that could perform a wide range of tasks at a human level, and the development of Orion is seen as a crucial step toward that ambitious target.
i love being home in the midwest.
the night sky is so beautiful.
excited for the winter constellations to rise soon; they are so great.
— Sam Altman (@sama) September 14, 2024
The Verge asked ChatGPT o1-preview to interpret Sam Altman’s cryptic post on X, where he mentioned being “excited for the winter constellations to rise soon.” According to ChatGPT, the message was hinting at the word Orion, which is also a prominent winter constellation. The model elaborated that the phrase “winter constellations” directly pointed to Orion, and even noted that if one rearranges certain letters in Altman’s message, it could spell out “ORION.”
We asked ChatGPT ourselves and it said:
“OpenAI’s anticipated AI model, codenamed “Orion,” is reportedly on track for a potential release by December 2024. This model, speculated to be the next evolution of GPT technology, could initially become available to select partners through a phased rollout. While Microsoft is preparing to support Orion via Azure, OpenAI’s CEO Sam Altman recently downplayed some of the speculation, labeling it “fake news” on X, suggesting that details on the release timeline might still be in flux. Orion, expected to excel in advanced reasoning and multi-modal processing, represents OpenAI’s response to competitive advancements in the AI space.”
Training Orion involved using data from another model, called o1, which OpenAI released in September. This model, codenamed “Strawberry,” played an important role in generating synthetic data that was used to enhance the training process for Orion, making it more robust and capable. An OpenAI executive’s cryptic post on X hinted at the upcoming release, referring to “winter constellations,” a possible nod to Orion—a well-known constellation visible in winter.
Despite the ambitious timeline, it’s noted that these plans may still shift, as is often the case with complex AI projects. Neither OpenAI nor Microsoft provided comments on this story, leaving some uncertainty around the exact details of the release.
This release comes during a time of transformation for OpenAI, following a $6.6 billion funding round that requires a change to a for-profit structure. The funding round is notable not only for its size but also for the implications it has on the company’s future direction and strategy. Additionally, key personnel including CTO Mira Murati and Chief Research Officer Bob McGrew have recently announced their departures, along with VP of post-training, Barret Zoph.
Now all eyes are on how OpenAI will manage this next phase of growth and innovation.
Featured image credit: Kerem Gülen/Ideogram
