Openai has slammed the brake on the release of the open-source AI model awaited by its curiosity, citing the need for more rigorous security checks before allowing developers to get a hand on it. The launch, originally before this summer and then the next week, has now been postponed indefinitely. Chatgipt-Mekar CEO Sam Altman on Friday broke the news in a post on X (East Twitter), saying the company needed more time to evaluate the potential risks of the model.

“We need time to run additional safety tests and review high -risk sectors. We are not yet sure how long we will take,” Altman wrote. “When we believe that the community will create great things with this model, once the weight is out, they cannot be pulled back. It is new to us and we want to correct it.”

This is not just an AI release. The upcoming open model of Openai has been billed as one of the most exciting technique launch of summer, which is with the beginning of the GPT 5’s Lump (and still mysterious). But unlike the GPT 5, which is expected to be tightly controlled, the open model was designed to be downloadable and fully used by developers without developers, the first for Openai over the years.

However, this freedom comes with a catch. By providing the developers for the underlying “weight” of the model, the main parameters that define its intelligence loses the risk of losing the openIAI risks to lose the risk of how it is used. This appears to be in front and center in the decision to hit the worry break.

The head of the VP and Open Model Project of AIDAN CLARK, Openai’s research,, further argued in his own post: “Capacity war, we think the model is unprecedented – but our bar is high for an open source model, and we need some more time to ensure that we are releasing a model that we pride with every axis.”

While developers from all over the world will now have to wait a little longer to test the most powerful open model of openiI, the company is promising that it will be worth waiting. Inner sources say the model is expected to rival O-Series’s argument skills-the family of the model who power the GPT 4O-and was designed to improve all the open-sources options currently available.

But the delay of openi can open the door for the contestants as well. A few hours before the announcement, Chinese Startup Moonshot AI unveiled its latest Heavyweight: Kimi 2, a large-scale one-trilian-parameter model. The initial benchmark suggests that Km K2 already excludes the GPT 4.1 of OpenAII from a range of coding and agent works, which extends bets for openi’s open offer.

The open-source AI arms race is heating up, with Zai of Google Deepmind, Anthropic and Elon Musk in their next-gene model. For Openai, this delay means that temporarily citing the spotlight to its rivals, a rare step for the company that took the AI boom with a chatgpt.

Nevertheless, Altman indicated on some “unexpected and quite amazing”, when he first revealed the initial delay of the model in June, many were surprised whether the openia is sitting on a groundbreaking capacity, it is not ready to do unlike.