Sam Altman Confirms Indefinite Delay of OpenAI’s Open-Source Model Over Safety

Highlights

OpenAI delays its much-anticipated open AI model indefinitely, citing safety evaluations and high-risk concerns, says CEO Sam Altman.

OpenAI has once again postponed the release of its eagerly awaited open-source AI model, citing the need for more in-depth safety reviews and risk evaluations. The launch, initially expected earlier this summer and later scheduled for next week, has now been delayed indefinitely, according to OpenAI CEO Sam Altman.

The announcement came via a post on X (formerly Twitter), where Altman explained that the company is proceeding cautiously due to the unprecedented nature of this release.

“We need time to run additional safety tests and review high-risk areas. We are not yet sure how long it will take us,” Altman wrote. “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.”

Unlike OpenAI’s more tightly guarded models like GPT-4 and the anticipated GPT-5, the upcoming open model was set to be fully downloadable and accessible without limitations. This level of openness would mark a significant shift for the company, which has historically kept its most powerful AI tools under strict control.

The model’s release was highly anticipated by developers and researchers worldwide, promising unrestricted access to the “weights” — the underlying parameters that govern how the AI functions. However, this transparency also introduces serious concerns about misuse, especially when the model’s intelligence rivals current state-of-the-art systems.

Aidan Clark, OpenAI’s VP of Research and the lead on the open model project, offered further insights into the company’s rationale.

“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 posted.

Despite the delay, internal sources suggest the model is designed to match — and possibly surpass — the reasoning capabilities of OpenAI’s o-series models, which include the powerful GPT-4o. Its potential to outperform current open-source alternatives has made it one of the most anticipated tech releases of 2025.

However, the indefinite delay gives competitors an opportunity to catch up — or even leap ahead. Notably, Chinese AI firm Moonshot AI unveiled its own mega-model, Kimi K2, just hours before OpenAI’s announcement. The trillion-parameter model reportedly outperforms GPT-4.1 on complex tasks, adding pressure to OpenAI’s cautious approach.

The open-source AI landscape is becoming increasingly competitive, with tech giants like Google DeepMind, Anthropic, and Elon Musk’s xAI investing heavily in their own cutting-edge models. OpenAI’s decision to pause — even temporarily — could shift momentum in a rapidly evolving arms race.

Still, hints from Altman suggest there may be more at play. Back in June, when the first delay was announced, he teased something “unexpected and quite amazing” tied to the model, fueling speculation that OpenAI may be holding back a major breakthrough.

Until then, developers will have to wait — but the promise remains that when the model is finally released, it will be worth it.