OpenAI has disputed recent reports about its product roadmap, saying it does not intend to release its AI model, codenamed “Orion,” this year.
“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 technology coming out.”
The Verge reported Thursday that Orion, expected to be OpenAI's next frontier model, will be available by December and trusted partners will be the first to preview it ahead of deployment through ChatGPT. According to The Verge, Microsoft, a close OpenAI collaborator and investor, expects to have access to Orion as early as November.
OpenAI previously told TechCrunch that The Verge's report was not accurate, but did not provide further details.
Orion, a step up from OpenAI's current flagship GPT-4o, is reportedly partially trained on synthetic training data from o1, the company's “inference” model. For the time being, OpenAI plans to continue developing its new “GPT” model in parallel with inference models like o1, which it believes can address fundamentally different use cases.
OpenAI's statement leaves considerable room for change. Maybe the company's next major model won't actually be an Orion. Alternatively, OpenAI may release a new model by December, but with less functionality than Orion.
At this point, it's anyone's guess.