OpenAI to Release First ‘Open’ Language Model Since GPT-2 in 2019

OpenAI is set to release its first “open-weight” language model in years, marking a significant departure from its more recent closed systems like GPT-3 and GPT-4. This new model, expected to arrive later in 2025, will allow developers to run, modify, and deploy the model on their own hardware. Unlike open-source models, an open-weight language model is publicly accessible for use but may have certain restrictions on modification or redistribution.

A Return to Openness

In an update posted to X (formerly Twitter) on March 31, OpenAI CEO Sam Altman shared that the company would launch this new open-weight model, designed to include reasoning capabilities, after gathering developer feedback.

Altman emphasized that while OpenAI has been considering this move for some time, other priorities had delayed the release. However, with the growing demand for accessible, customizable AI tools, the decision to move forward with the model release now felt particularly important.

“This will be the first ‘open-weight’ model since GPT-2 in 2019,” Altman explained. GPT-2, OpenAI’s previous open-weight model, had a staggered release, with a partial release in February 2019 followed by a full release in November of the same year.

Developer Events and Feedback

To ensure the model meets the needs of developers, OpenAI will host a series of developer events starting in San Francisco, California, in the coming weeks. These events will be followed by sessions in Europe and the Asia-Pacific region, where developers will be able to explore early prototypes and share their feedback on how the model can be improved for future use.

Altman noted that OpenAI was particularly excited to see how developers would utilize the new open-weight model and how large companies and governments might adapt it to their own infrastructures. OpenAI plans to take additional measures to ensure the model is secure and adaptable, given the expectation that the model will be modified after its release.

A Shift in AI Landscape

The release of this open-weight model comes at a time when the AI landscape is rapidly evolving. Competitors like DeepSeek, a new AI model that functions similarly to ChatGPT but at a lower cost, are intensifying the competition. Additionally, companies like Alibaba and Google have introduced new open-source models like Gemini 2.5 and cost-effective AI agents to challenge OpenAI’s dominance in the space.

Furthermore, Meta has reached a milestone with its AI model family Llama, which surpassed 1 billion downloads in a year since its release. This intense competition is fueling an AI arms race, which makes OpenAI’s decision to return to a more open approach to model release particularly timely.

Looking Ahead

Altman’s announcement comes just weeks after the CEO hinted at the upcoming release of GPT-4.5 and GPT-5, expected to arrive in the near future. As OpenAI prepares to release this new open-weight model, the company’s ongoing focus on expanding its offerings while keeping the development process transparent is seen as a key move in fostering innovation in the AI space. This release could spark new opportunities for businesses, researchers, and developers, particularly those seeking more control over their AI models and workflows.

As OpenAI moves forward with its open-weight language model, the industry will be watching closely to see how this development impacts the growing field of AI and its many applications.