OpenAI is delaying the release of its open model again, extending the wait for developers who expected to test it next week. CEO Sam Altman said Friday that the company needs more time for safety work before making the model available.
The decision matters because this is not a typical product update. Once an open model’s weights are released, OpenAI cannot simply take them back from the public internet or from developers running the model locally.
Why OpenAI is holding back the model
Altman said the release is being pushed back indefinitely so OpenAI can run additional safety tests and review high-risk areas. The company had already delayed the model by a month earlier this summer, and it is now avoiding a new firm timeline.
we need time to run additional safety tests and review high-risk areas. we are not yet sure how long it will take us
In the same post on X, Altman framed the issue as a release decision with lasting consequences. OpenAI believes the community will build valuable projects with the model, but the company is treating the publication of weights as a point of no return.
while we trust the community will build great things with this model, once weights are out, they can’t be pulled back. this is new for us and we want to get it right.
That is the central tension around the OpenAI open model. A downloadable model gives developers more freedom, including the ability to run it locally. But that same freedom reduces the company’s ability to control what happens after release.
What developers are waiting for
The delayed model is one of the most anticipated AI releases of the summer, alongside OpenAI’s expected release of GPT-5. Unlike GPT-5, the open model is expected to be available for developers to freely download and run locally.
That makes it important for developers who want direct access to model weights rather than relying only on cloud-hosted systems. Local execution can change how developers test, build, and deploy AI tools, because the model is not limited to a hosted interface controlled by the company.
TechCrunch previously reported that the model is expected to have reasoning capabilities similar to OpenAI’s o-series of models. It also reported that OpenAI planned for the model to be best-in-class compared with other open models.
Those expectations raise the stakes of the delay. If OpenAI releases a strong open AI model, it could give developers a major new option in the open model ecosystem. If it waits too long, competitors may keep filling the space with their own releases.
The competitive pressure is rising
OpenAI is trying to show that it remains Silicon Valley’s leading AI lab. That task is getting harder as xAI, Google DeepMind, and Anthropic invest billions of dollars in their own AI efforts.
The open model market also became more competitive this week. Earlier on Friday, Chinese AI startup Moonshot AI launched Kimi K2, a one-trillion-parameter open AI model. According to the source article, Kimi K2 outperforms OpenAI’s GPT-4.1 AI model on several agentic-coding benchmarks.
That launch adds pressure because open models compete not only on access, but also on capability. Developers evaluating AI models often care about whether a model can handle reasoning, coding, and complex workflows well enough to justify adopting it.
For OpenAI, the delay creates a difficult balance. The company wants to release something strong enough to meet developer expectations, but it also wants to avoid publishing a model before it has completed the safety work it believes is necessary.
What OpenAI has said about quality
In June, Altman said OpenAI had achieved something unexpected and quite amazing with the model, but did not explain what he meant. That comment increased interest around the release without giving developers many concrete details.
Aidan Clark, OpenAI’s VP of research who is leading the open model team, also emphasized that the company is setting a high standard for the release.
Capability wise, we think the model is phenomenal — but our bar for an open source model is high and we think we need some more time to make sure we’re releasing a model we’re proud of along every axis,
Clark’s statement points to a release that is being judged on more than raw capability. Based on OpenAI’s comments, the company is weighing safety, quality, and readiness before putting the model into developers’ hands.
There are still open questions about what will make it into the final version. TechCrunch previously reported that OpenAI leaders have discussed enabling the open AI model to connect to the company’s cloud-hosted AI models for complex queries. The source article says it is unclear whether those features will be included in the final open model.
What the delay means now
For now, developers have no new release date. OpenAI has moved from a planned release next week to an indefinite delay while it continues safety testing.
The practical result is simple: anyone waiting to download and run OpenAI’s first open model in years will have to wait longer. The broader implication is that OpenAI is treating this launch as a strategic and safety-sensitive moment, not just another model update.
The company’s final decision will shape how it participates in the open AI model ecosystem. Until then, attention will stay on two questions: when OpenAI releases the model, and whether it can meet the high expectations the company has helped create.