OpenAI Delays Launch of Its Open Model Again

On Friday, Sam Altman, the CEO of OpenAI, announced that the launch of the company’s open model will be postponed. This follows a prior delay of one month earlier this summer. Initially set for next week, Altman confirmed that the release is now indefinitely postponed for further safety evaluations.

“We need more time to conduct safety assessments and evaluate high-risk aspects. The timeline for this is currently uncertain,” Altman mentioned in a post on X. “While we are confident in the community’s capability to develop remarkable projects with this model, once the weights are public, they cannot be retracted. This is a new frontier for us, and we want to ensure everything is executed properly.”

The expected release of OpenAI’s open model is one of the most anticipated AI highlights of the summer, alongside the upcoming launch of GPT-5. Unlike GPT-5, OpenAI’s open model will be freely downloadable, allowing developers to run it locally. With both launches, OpenAI seeks to reaffirm its status as the leading AI lab in Silicon Valley—a challenging task as xAI, Google DeepMind, and Anthropic invest significantly in their own ventures.

This delay means developers will have to wait longer to utilize the first open model OpenAI has released in years. According to TechCrunch, OpenAI’s open model is expected to feature reasoning capabilities akin to its o-series models and aim to be industry-leading compared to other open offerings.

The competition in the open AI model space escalated this week. Earlier on Friday, the Chinese AI startup Moonshot AI debuted Kimi K2, a one trillion parameter open AI model that outperforms OpenAI’s GPT-4.1 in several agentic-coding metrics.

In June, when Altman first announced the delays concerning OpenAI’s open model, he hinted at achieving something “unexpected and quite amazing,” though he did not elaborate further.

“We believe the model is outstanding in terms of capability—but our expectations for an open-source model are high, and we need more time to ensure we’re releasing a model that meets our standards comprehensively,” stated Aidan Clark, OpenAI’s VP of research, who leads the open model team, in a post on X Friday.

TechCrunch previously reported that OpenAI officials are contemplating the possibility of enabling the open AI model to connect with the company’s cloud-hosted AI models for more sophisticated queries. However, it is still uncertain whether these features will be integrated into the final release of the open model.