In an unexpected move, OpenAI has announced a delay for the launch of its open-source model, which the developer community had been eagerly anticipating for the coming week.
Sam Altman, the company's CEO, explained the decision was made to give the team more time for additional safety testing and to review high-risk areas, without setting a new launch date.
Altman's announcement came via a post on the X platform, where he noted the sensitive nature of the release.
He stated in his post, "we need time for additional safety testing and to red-team high-risk areas. not sure how long it will take."
He added that the company wants to be completely sure of the model's integrity, as retracting its weights after a public release is not feasible.
This marks the second such delay for this model. The company previously announced an initial postponement back in June.
At that time, Altman attributed the move to an "unexpected and quite amazing" breakthrough in the team's research, without offering further details.
In this context, Aidan Clark, OpenAI's VP of Research, supported the latest delay. Clark stated on X that the model's current capabilities are technically "immense," but the company's standards for releasing an open-source model are very high. He affirmed the team needs more time to ship a product they can be proud of in all respects.
The decision comes as competition in the AI space intensifies. Chinese firm Moonshot AI, for instance, recently launched its open Kimi K2 model, which has surpassed GPT-4.1's performance in some benchmarks.
Meanwhile, companies like Google, Anthropic, and xAI continue their massive investments to develop their own models, placing significant pressure on OpenAI to maintain its leading position.
The open-source model was intended to allow developers to download and run it on their own hardware, offering greater flexibility and customization.
Altman had previously admitted that his company was on the "wrong side of history" regarding the open-source debate, framing the upcoming model as a response to strong demand from developers and enterprise clients who prefer open models for their flexibility.
With this postponement, developers and specialists will have to wait longer to experience OpenAI's first open-source model in years, as the company continues its cautious approach to ensuring its technology is safe before placing it in everyone's hands.
