OpenAI says it has no intention of releasing an artificial intelligence model codenamed Orion this year, contradicting recent reports about the company’s product development plan.
“We have no plans to release a model codenamed Orion this year,” a spokesperson told TechCrunch via email. “We have a lot of other great technologies coming out.”
On Thursday, The Verge reported that Orion, expected to be OpenAI’s next pioneering model, will be launched in December and that trusted partners will be the first to be able to preview it before deployment via ChatGPT. According to The Verge, Microsoft, a close collaborator and investor in OpenAI, expects to gain access to Orion as early as November.
OpenAI previously told TechCrunch that The Verge’s report was not true, but declined to elaborate.
A step up from OpenAI’s current flagship, GPT-4o, Orion is apparently trained partially on synthetic training data from o1, the company’s “inference” model. OpenAI plans to continue developing recent “GPT” models in the near future, along with inference models such as o1, which it says address fundamentally different exploit cases.
OpenAI’s statement leaves considerable room for maneuver. It may be that the company’s next major model won’t actually be Orion. Or maybe OpenAI will release a recent model by December, but it will be less proficient than Orion.
You can guess it at this point.