Sam Altman says reports of OpenAI’s next frontier model are “fake news out of control” as speculations of a December launch mount

Sam Altman says reports of OpenAI's next frontier model are "fake news out of control" as speculations of a December launch mount

What you need to know

  • A new report suggests OpenAI is gearing up to ship a next-gen frontier AI model code-named Orion in December.
  • The report further suggested that the ChatGPT maker would give companies it works closely with priority access to the model, including Microsoft.
  • Sam Altman responded to the claims on X, branding it as “fake news out of control.”

As someone who has closely followed the AI industry for years and witnessed its rapid evolution, I must admit that the recent developments surrounding OpenAI have left me quite intrigued and a bit bewildered at times.


Over the past few months, OpenAI has faced several challenges, such as the departure of key executives, financial difficulties, and allegations that it’s violating U.S. copyright law. Notably, a new lineup of cutting-edge AI models from OpenAI, including the OpenAI o1 and o1-mini, were introduced about a month ago. These models boast advanced problem-solving capabilities in coding, mathematics, and science.

The Strawberry AI models are proving to be highly promising according to the benchmarks presented by OpenAI. A recent investigation found that OpenAI’s model o1 could successfully pass OpenAI’s coding interview for research engineers at a rate of 90-100%, which might stir unease among software developers regarding their employment prospects in an AI-dominated future.

According to a recent report by The Verge, the creators of ChatGPT are planning to release their latest model, known as Orion, in December, marking two years since the launch of ChatGPT. Unlike previous releases, this new AI model won’t be made available to the general public through ChatGPT. Instead, the company is focusing on providing access to select companies it collaborates with closely, such as Microsoft, who may host the model on their Azure platform as early as November.

Based on information from The Verge, Orion is considered as the potential successor to OpenAI’s GPT-4. It remains undecided if OpenAI will market Orion as GPT-5, following speculation that the company might be shifting away from the conventional naming scheme for its models. Nevertheless, Sam Altman hinted that the successor of GPT-4 would exhibit greater intelligence and operate similar to a “synthetic brain.

It was all a hoax?

Sam Altman says reports of OpenAI's next frontier model are "fake news out of control" as speculations of a December launch mount

After coming across The Verge’s report about the anticipated launch of a new-generation AI model, its widespread circulation was reminiscent of a wildfire. However, recent developments suggest that we may need to approach this information with caution. This is particularly so following a statement made by OpenAI CEO Sam Altman, who labeled the report as “uncontrolled fake news.

fake news out of controlOctober 25, 2024

This doesn’t necessarily imply that The Verge’s report is incorrect. It could be a question of timing or some other factor at play. You might recall that on previous occasions, we had anticipated OpenAI to deliver a specific product or service by a certain date, but it often turned out to be something else entirely.

Back in May this year, whispers suggested that OpenAI was preparing to debut an AI-driven search rival to Google, along with GPT-5. However, during their special event, OpenAI’s CEO Sam Altman clarified on social media that the company wasn’t releasing an AI engine or GPT-5 as initially speculated. Instead, he teased that the event would be extraordinary and alluded to the launch of GPT-40.

It turned out that during the development of their product, OpenAI put significant stress on its safety team, hastening the testing process to just a week. During this period, several key executives left the AI firm, with some claiming that safety concerns were overlooked in favor of quickly releasing attractive products. However, an OpenAI representative acknowledged the haste of the launch and sending out invitations before thoroughly testing the model, but emphasized that no corners were cut in the process.

Read More

2024-10-25 14:39