OpenAI delays the release of its open model, again

In a recent announcement, OpenAI CEO Sam Altman revealed that the company is delaying the release of its eagerly awaited open model. Initially slated for launch next week, the release has now been postponed indefinitely to allow for more comprehensive safety evaluations. Altman emphasized the importance of thorough testing, stating, "We need time to run additional safety tests and review high-risk areas. We are not yet sure how long this will take." He acknowledged the community's eagerness to innovate with the new model, but underscored the irreversible nature of releasing such technology. The anticipation surrounding OpenAI's open model is palpable, particularly as it stands alongside the expected launch of GPT-5. Unlike GPT-5, this open model will be available for developers to download and operate locally, marking a significant step in OpenAI's mission to remain a leader in the competitive AI landscape, especially with other players like xAI, Google DeepMind, and Anthropic making substantial investments. The delay means developers will have to bide their time before accessing the first open model from OpenAI in years. According to prior reports, this open model is expected to possess reasoning capabilities comparable to the company's existing o-series models and aims to be the best in its class among open models. The competition in the realm of open AI models intensified this week with the launch of Kimi K2 by Chinese startup Moonshot AI, which boasts one trillion parameters and surpasses OpenAI’s GPT-4.1 in various coding benchmarks. In June, Altman hinted at unexpected advancements during earlier delays but did not provide specifics. Aidan Clark, OpenAI's VP of research, also shared insights, stating, "Capability-wise, we think the model is phenomenal — but our bar for an open-source model is high, and we believe we need more time to ensure we’re releasing a model we’re proud of along every axis." While discussions have occurred regarding the potential for the open model to integrate with OpenAI’s cloud-hosted AI models for complex queries, it remains uncertain if these features will be included in the final version.