Citing the necessity for additional safety testing, OpenAI has once again postponed the release of its eagerly awaited open AI model.
Only a few weeks have passed since the corporation postponed the release date, which was initially planned for earlier this summer.
In a post on X (formerly Twitter) on Friday, OpenAI CEO Sam Altman confirmed the delay and stated that the model would not be distributed next week as planned. “We need time to analyse high-risk locations and conduct additional safety testing. How long it will take us is still unknown to us,” Altman said. Weights cannot be taken back once they are out, even though we have faith that the community will use this concept to create amazing things.
Since this is new to us, we want to do it correctly.
In contrast to OpenAI’s customary proprietary strategy, the impending release was to be the company’s first publicly available AI model in years and was anticipated to be available for developers to download and run locally.
The delay occurs as the open-source AI market is becoming more competitive. The same day, Moonshot AI, a Chinese AI startup, announced the release of its Kimi K2 model, a one-trillion-parameter system that supposedly beats OpenAI’s GPT-4.1 on a number of sophisticated coding tests.
Aidan Clark, Vice President of Research at OpenAI, who leads the open model team, also offered his thoughts on the hold-up. Clark wrote on X, “We think the model is fantastic in terms of capability, but we have high expectations for an open source model, and we believe we need some more time to make sure we’re releasing a model we’re proud of along every axis.”
In terms of reasoning skills, industry observers had anticipated that OpenAI’s open model would be comparable to its O-series models. Expectations among developers and researchers were further raised when TechCrunch previously reported that the company was seeking to create a best-in-class open model.
The delay highlights the difficulties even top labs have in striking a balance between innovation and safety and ethical issues, even though it is unlikely to lessen OpenAI’s reputation in the artificial intelligence space.
With companies like xAI, Google DeepMind, and Anthropic all making significant investments in next-generation models, the AI environment has grown more congested in recent months.
During the announcement of the new model’s initial postponement in June, Altman alluded to its significance by saying the team had accomplished something “unexpected and quite amazing,” but he did not elaborate.
