OpenAI Set to Launch Its Next Major AI Model by December

OpenAI’s latest project, Orion, is on track to redefine the AI landscape, with a launch potentially happening by December, according to *The Verge*. Unlike OpenAI’s past models, Orion’s initial release won’t be made available to the public through ChatGPT. Instead, OpenAI plans to partner exclusively with select companies, granting them early access to the model’s capabilities. This approach allows these organizations to build innovative, customized tools and applications with Orion’s advanced power, underscoring OpenAI’s shift towards industry-focused deployment strategies, as shared by an insider familiar with the company's plans.

Microsoft, OpenAI’s strategic partner, is also involved in the launch preparations. According to a second source, Microsoft’s engineering team is working to deploy Orion on its Azure cloud platform as early as November, making it accessible to enterprise users through the cloud. While Orion is poised as OpenAI’s most advanced successor to GPT-4, there’s still some uncertainty about the official naming convention; the company hasn’t confirmed if the model will be marketed as GPT-5. Given the fast-paced nature of AI development, OpenAI’s release timelines are always subject to potential adjustments, and unforeseen changes could still occur. At present, however, neither OpenAI nor Microsoft has commented publicly on these plans.

Orion is set apart by an unprecedented scale of power and capability. An OpenAI executive has hinted that it may be up to 100 times more powerful than GPT-4, placing it far beyond the recently launched o1 model, which was designed primarily to improve reasoning. OpenAI’s longer-term vision with Orion and future models is to combine various model capabilities to create an even more robust and versatile AI system, ultimately striving toward an ambitious goal: artificial general intelligence (AGI). For OpenAI, Orion represents an important leap forward in that direction.

Interestingly, OpenAI reportedly used its o1 model, internally known as *Strawberry*, to create synthetic data for Orion’s training, a novel method to improve the model’s performance. Following the training phase, OpenAI’s team celebrated with a casual gathering in September, as recounted by a source close to the matter. Around the same time, OpenAI’s CEO, Sam Altman, posted a cryptic message on social media, referencing “winter constellations rising soon.” The timing of this post led to speculation that Altman was hinting at the impending arrival of Orion, named after the prominent constellation visible in the winter sky from November through February. ChatGPT o1-preview responded amusingly to questions about the post, pointing out the connection to Orion and attempting to rearrange letters to form the name—a playful nod to the community’s anticipation.

Orion’s launch comes during a period of significant transformation for OpenAI. The company recently secured a groundbreaking $6.6 billion in funding, a financial milestone that requires OpenAI to reconfigure as a for-profit entity. This structural shift marks a new chapter in OpenAI’s journey, blending its commitment to advancing AI with practical financial goals and a growing emphasis on productization and commercialization. However, this period of expansion is also marked by significant leadership changes: Mira Murati, the company’s Chief Technology Officer, recently announced her departure, followed closely by Chief Research Officer Bob McGrew and VP of Post Training Barret Zoph. These departures reflect a restructuring within OpenAI, as it navigates the complexities of rapid growth while maintaining its core research and development focus.

Through Orion, OpenAI is taking bold steps to make AI more powerful and useful, aligning with its vision of creating transformative technology. As the next chapter unfolds, this model release will not only test OpenAI’s technical and strategic ambitions but will also influence the broader landscape of artificial intelligence.