OpenAI says it doesn’t plan to release an AI model codenamed Orion this 12 months, contradicting recent reports concerning the company's product roadmap.
“We don’t have any plans to launch a model codenamed Orion this 12 months,” a spokesperson told TechCrunch via email. “We plan to bring many other great technologies to market.”
The Verge reported Thursday that Orion, expected to be OpenAI's next top-of-the-line model, would launch in December and that trusted partners could be the primary to preview it ahead of a launch via ChatGPT. According to The Verge, Microsoft, an in depth OpenAI collaborator and investor, expects to achieve access to Orion as early as November.
OpenAI previously told TechCrunch that The Verge's report was incorrect, but declined to elaborate.
Orion, an evolution of OpenAI's current flagship, GPT-4o, is allegedly partially trained using synthetic training data from o1, the corporate’s “reasoning” model. For the foreseeable future, OpenAI plans to develop latest “GPT” models alongside reasoning models like o1, which it believes cover fundamentally different use cases.
OpenAI's statement leaves considerable scope. It may very well be that the corporate's next big model isn't Orion. Or perhaps OpenAI will release a brand new model by December, but it’ll be less powerful than Orion.
At this point, it's anyone's guess.
