published an exclusive and seemingly well-researched and sourced report last night (in my view it's great, read it here) by journalists Kylie Robison and Tom Warren, which say that OpenAI plans to launch one other recent frontier AI model codenamed Orion by December – which can or is probably not GPT-5.
But two hours after the article went online, Sam Altman, co-founder and CEO of OpenAI, selected X to reply by directly responding to Robison's part within the article and writing “Fake news uncontrolled.”
Altman hasn't elaborated much since then on what I saw, and the response notably isn't exactly a direct denial of the claims – he hasn't written “no” or “that's unsuitable,” let alone detailed the article is fake : Isn't OpenAI working on a brand new frontier model called Orion? That would contradict prior reporting from outlets, including that it actually makes such an effort internally – which, so far as I do know, OpenAI has never directly denied. Isn't it scheduled for release later this yr?
But it’s clearly an try to roll back reporting in its current form.
Given the accuracy of the report, it's an interesting quasi-denial, mentioning specific details about Orion's alleged release plans and the incontrovertible fact that it appears to be geared toward enterprise customers and will initially only be made available via an application programming interface (API). :
“
OpenAI's last release of a brand new Frontier model – o1 Preview and o1-mini – was in early September, just over a month ago. However, wider adoption of those large language models (LLMs) has been largely muted, partially because they’re expensive to operate for each the business and developers, and in addition because they’ve a brand new “logical” architecture and are in lots of cases more limited far faster than OpenAI's GPT family of models and can’t currently accept file uploads or generate and analyze images.
A brand new frontier model would help OpenAI once more stand out from competitors like Anthropic, which just this week unveiled a promising recent agent mode called Computer Use and a new edition of its Claude LLM family. OpenAI just isn’t planned
Whether or not OpenAI releases a brand new frontier model later this yr, we will probably be watching closely. Fans of the corporate and its models shouldn't have too high hopes in the interim.
