Join our daily and weekly newsletters to receive the latest updates and exclusive content on our industry-leading AI coverage. Find out more
Edge an exclusive and apparently well-researched and sourced report was released last night (I think it’s great, read it here) from journalists Kylie Robison and Tom Warren stating that OpenAI plans to release another up-to-date, pioneering AI model codenamed Orion – which may or may not be GPT-5 – by December.
However, two hours after the article was published Sam Altman, co-founder and CEO of OpenAI, took on X respond by responding directly to part of Robison’s article, writing “fake news is out of control.”
From what I’ve seen, Altman hasn’t elaborated much since then, and the response is notably not a direct denial of the claims – he didn’t write “No” or “this is false,” much less describe which part detailed article is wrong: isn’t OpenAI working on a up-to-date boundary model called Orion? That would be contradictory previous outlet reports incl Information that it is indeed making such efforts internally – which, to my knowledge, OpenAI has never directly denied. Isn’t the premiere planned for this year?
However, this is a clear attempt to reject the current reporting.
This is an compelling quasi-denial, given its precision Edge the report details Orion’s alleged release plans and the fact that it appears to be aimed at enterprise customers and will likely be initially made available via an application programming interface (API):
“Unlike the last two OpenAI models, GPT-4o AND o1Orion will not be widely available via ChatGPT initially. Instead, OpenAI plans to first grant access to companies it works closely with so they can build their own products and features, according to a source familiar with the plan.
Another source tells The Verge that engineers from Microsoft – OpenAI’s main partner in implementing artificial intelligence models – are preparing to host Orion on Azure in November. While Orion is seen within OpenAI as the successor to GPT-4, it is unclear whether the company will call it GPT-5 externally.“
The last version of OpenAI’s up-to-date frontier model – o1 Preview and o1-mini – arrived in early September, just over a month ago. However, wider reception of these huge language models (LLMs) has been largely muted, in part because they are exorbitant to operate for both the company and developers, and also because they have a up-to-date “reasoning” architecture and are in many cases more confined way than OpenAI’s GPT family of models, currently unable to accept file uploads or generate and analyze images.
The up-to-date pioneering model would facilitate OpenAI get back into the spotlight among rivals, including Anthropic, which just this week unveiled a promising up-to-date agent mode called “Computer Usage” and a up-to-date version of the LLM Claude family. OpenAI is not in the second lieutenant
Whether OpenAI releases a pioneering up-to-date model later this year or not, we’ll be following closely. For now, it seems that fans of the company and its models shouldn’t get their hopes up too early.