OpenAI has delayed an open-source model it had previously announced would release in early summer until later in the year. CEO Sam Altman wrote on X that the postponement is because the research team made an unexpected breakthrough – one he described as “amazing” and worth the wait – but noted that it will not be ready in June and will instead be launched in the summer months
The upcoming open model is projected to be as reasonable as the closed-source o-series models developed by OpenAI, and should surpass its competitors such as the R1 developed by DeepSeek. As other labs put out their own enhanced versions, with Mistral putting out its Magistral line this very week and China-based Qwen putting out hybrid reasoning models in April, the pressure is on OpenAI to produce a version that really shines
To remain competitive, OpenAI is also thinking of packaging more capabilities with the new model, including the ability to connect to its own cloud-based systems in order to process complex queries. These extras are yet to be determined though whether they shall make it to the final version or not
Open-sourcing a high-performing model is a big step for OpenAI, which has published only proprietary versions in years. Altman has received criticism about restricting access to the powerful features of OpenAI, and has said that openness had placed the company on the “wrong side of history”. The release would assist in mending ties with researchers and developers who have demanded greater openness
This delay implies a few things: OpenAI is actively working on improving its open model to make it live up to high expectations, be benchmark-competitive, and potentially include innovative elements. It is also indicative of the changing competitiveness of the AI sphere–in which research labs such as Mistral and Qwen are making inroads in open-model development.
The hyped release is not just a matter of code, but a message. In case the open model works as promised, OpenAI might establish itself as a leader in transparent AI development. With rivalry in the open-model space getting intensified, how the business strikes the balance between openness and performance could well be the theme of AI developments going forward.
In the meantime, developers and researchers will need to wait till later this summer to get access to the model but with the assurance that it would be something “worth the wait.”
