OpenAI has announced its intention to release its first “open” language model since GPT‑2 in the upcoming months. This information was shared through a feedback form published on the company’s website on Monday. OpenAI is inviting “developers, researchers, and members of the broader community” to provide their input through the form, which includes questions such as, “What would you like to see in an open-weight model from OpenAI?” and “What open models have you used previously?”
“We’re eager to collaborate with developers, researchers, and the wider community to gather insights and enhance the utility of this model,” OpenAI stated on its website. They also invited individuals interested in participating in feedback sessions with the OpenAI team to express their interest in the form. To collect feedback, OpenAI plans to host developer events, with the first set to occur in San Francisco within a few weeks, followed by additional sessions in Europe and the Asia-Pacific regions. The move comes as OpenAI faces growing competition from rivals like the Chinese AI lab DeepSeek, which have embraced an “open” approach to model releases. These competitors provide their models to the AI community for experimentation and potential commercialization, which has proven to be an effective strategy for some organizations. For instance, Meta reported earlier in March that its Llama family of open AI models had surpassed 1 billion downloads, while DeepSeek has rapidly garnered a significant global user base and attracted interest from domestic investors.
In a recent Q&A on Reddit, OpenAI CEO Sam Altman acknowledged that the organization may have taken the wrong approach regarding open-sourcing its technologies. “I personally think we need to explore a different open source strategy,” Altman remarked. “Not everyone at OpenAI shares this perspective, and it’s not our current top priority… We will create better models moving forward, yet we will maintain a smaller lead than we did in previous years.” Altman further elaborated on OpenAI’s plans for an open model in a post on X on Monday afternoon, indicating that this upcoming model will possess reasoning capabilities similar to OpenAI’s o3-mini. “Before release, we will assess this model in accordance with our preparedness framework, just as we would for any other model,” he noted. “We will also undertake additional work considering that this model will likely be modified post-release… We’re excited to see what developers will create and how large companies and governments might utilize it when they prefer to run a model independently.”
Additionally, excerpts from an upcoming book by Wall Street Journal reporter Keach Hagey, released over the weekend, claim that Altman misled OpenAI executives regarding model safety reviews in the past.
