ChatGPT Upgrade Orion Model Launching In December? What Sam Altman Said – News18

Last Updated:October 29, 2024, 08:16 IST

OpenAI’s next AI model could be 100 times faster than ChatGPT which will surely worry its rivals

Is OpenAI going to announce its new AI model, Orion in December this year? According to a report byThe Verge, the Microsoft-backed company, which is behind ChatGPT, is preparing to introduce Orion before the year ends, but this launch may differ from previous models.

Instead of being released widely through ChatGPT, Orion’s initial rollout will be limited to partner companies, including Microsoft, which will leverage the AI model to enhance their products and features, as noted in the report.

However, Sam Altman, CEO and co-founder of OpenAI, has dismissed the December 2024 timeline, referring to the report as “fake news out of control.” Altman shared his thoughts on X (formerly Twitter), responding to the claim about Orion’s December release without explicitly addressing the details mentioned in the article.

Soon, in a subsequent post, Altman stated that there will be “plenty of great stuff coming your way,” and added, “just offends me how media is willing to print random fantasy.”

Following Sam Altman’s dismissal of the article as “fake news,” OpenAI spokesperson Niko Felix told The Verge that the company does not have plans to release a model code-named Orion this year. But Felix noted that OpenAI is focused on unveiling other innovative technologies in the near future.

As per The Verge, Microsoft, a significant investor in OpenAI, intends to include the Orion AI model in its Azure services as early as November. Still unknown, though, is whether OpenAI would stick with the Orion moniker in the public domain or switch to the more well-known GPT-5 name in the end.

In March 2023, GPT-4 was introduced. Since then, the company has released GPT-4o and, more recently, GPT-o1, codenamed Strawberry. It is claimed that Orion, or whatever OpenAI decides to call it externally, is 100 times more powerful than GPT-4.

It is not to be confused with models such as GPT-o1 ‘Strawberry’ or GPT-4o, though. Orion might reportedly have been trained with OpenAI o1 data, which would have improved its capacity for more complex reasoning and handling of bigger requests. However, unlike Meta’s Llama family of AI models, OpenAI’s Orion is probably going to stay a closed model, which means it won’t be publicly accessible.