Be part of our every day and weekly newsletters for the newest updates and unique content material on industry-leading AI protection. Be taught Extra
The Verge final night time printed an unique and seemingly properly researched and sourced report (it’s nice in my view, learn it right here) from journalists Kylie Robison and Tom Warren stating that OpenAI plans to launch one other new frontier AI mannequin, codenamed Orion — which can or might not be GPT-5 — by December.
But two hours after the article went reside, Sam Altman, OpenAI’s co-founder and CEO, took to X to reply by replying on to Robison’s share of the article, writing “pretend information uncontrolled.”
Altman hasn’t elaborated a lot since then from what I’ve seen, and the response is notably not precisely a direct denial of the claims — he didn’t write “No” or “that is false,” a lot much less describe which half of the detailed article is incorrect: is OpenAI not engaged on a brand new frontier mannequin known as Orion? That might contradict prior reporting from shops together with The Info that it does have such an effort internally — which to my information, OpenAI by no means straight denied. Is it not planning to launch later this yr?
However it’s clearly an try to push again on the reporting because it stands.
It’s an fascinating quasi-denial given how exact The Verge report is, noting particular particulars about Orion’s supposed launch plans and the truth that it seems to be geared towards enterprise clients and presumably can be served up by means of an utility programming interface (API) solely at first:
“Not like the discharge of OpenAI’s final two fashions, GPT-4o and o1, Orion gained’t initially be launched broadly by means of ChatGPT. As an alternative, OpenAI is planning to grant entry first to firms it really works carefully with to ensure that them to construct their very own merchandise and options, based on a supply aware of the plan.
One other supply tells The Verge that engineers inside Microsoft — OpenAI’s principal accomplice for deploying AI fashions — are getting ready to host Orion on Azure as early as November. Whereas Orion is seen inside OpenAI because the successor to GPT-4, it’s unclear if the corporate will name it GPT-5 externally.“
OpenAI’s final launch of a brand new frontier mannequin — o1 preview and o1-mini — occurred in early September, a bit greater than a month in the past. But the broader reception of those giant language fashions (LLMs) has been largely muted, partly as a result of they’re costly for each the corporate and builders to function, and likewise as a result of they’re of a brand new “reasoning” structure and are extra restricted in some ways than OpenAI’s GPT household of fashions, unable at the moment to simply accept file uploads, or to generate and analyze imagery.
A brand new frontier mannequin would assist OpenAI seize the limelight once more from rivals together with Anthropic, who simply this week unveiled a promising new agentic mode known as “Laptop Use” and new model of its Claude household of LLMs. OpenAI isn’t in ppor
Whether or not OpenAI does find yourself releasing a brand new frontier mannequin later this yr or not, we’ll be following carefully. For now, it appears, followers of the corporate and its fashions shouldn’t get their hopes up too quickly.