OpenAI has announced a delay in the release of its highly anticipated open-weight AI model, which was originally expected to launch in early summer 2025. The announcement came from CEO Sam Altman on June 11 via a post on X (formerly Twitter), where he stated that the company needs “a little more time” and the release would occur later in the summer, but not in June.

Altman explained that the delay is due to remarkable, unforeseen progress made by the company’s research team. According to him, this unexpected breakthrough is significant enough to warrant a longer development period to fully realize its potential. He expressed confidence that the final outcome would be “very, very worth the wait.”

This marks OpenAI’s first open-weight model since GPT-2, which was released in 2019. Since then, the organization has shifted to developing and releasing closed-source models such as GPT-4.1 and the o3 and o4 model series. These newer models have not had their trained weights released publicly, making them accessible only through OpenAI’s proprietary platforms.

The forthcoming open-weight model is expected to showcase advanced reasoning capabilities, potentially matching the performance of the current o-series models. In fact, OpenAI claims it aims to surpass other prominent open-weight models like DeepSeek’s R1 in reasoning tasks. OpenAI is reportedly working on integrating additional sophisticated features to enhance this model further. For instance, TechCrunch reports that one possibility includes integrating the model with OpenAI’s cloud-based infrastructure to support more complex and demanding tasks. However, whether this capability will be included in the final release remains uncertain.

The delay and development of this new model are being closely watched in the tech and AI research community, particularly because open-weight models have a significant impact on democratizing access to advanced AI tools. Open-weight language models are artificial intelligence systems in which the trained parameters—the numerical data that determines how the model functions—are made publicly available. This allows developers, researchers, and institutions to run the models on their own hardware, without requiring API access to the host company’s cloud services.

While open-weight models allow greater flexibility and transparency, they are still typically governed by licensing agreements. These may limit how the models can be modified, redistributed, or used for commercial purposes. Nonetheless, the availability of such models contributes to wider experimentation, innovation, and accountability within the AI field.

OpenAI’s move to reintroduce an open-weight model after years of closed releases suggests a strategic shift or an effort to balance open research with commercial interests. The upcoming release is expected to generate significant interest, especially given OpenAI’s prominence and the promise of performance comparable to its best proprietary models. Until the release later this summer, the AI community will continue to speculate on the model’s capabilities and potential impact.