OpenAI to Launch Its First “Open-Weight” Language Model Since GPT-2 in 2019

OpenAI is preparing to launch an “open” version of its language model this year, allowing developers to run it on their hardware. CEO Sam Altman shared the update on March 31 via X, announcing that the new “open-weight” language model would be available in the coming months. OpenAI aims to gather feedback on how to make it as useful as possible before the release.

Altman mentioned, “We’ve been planning this for a while, but other priorities had to come first. Now, it feels important to move forward.” This model will be the first open-weight release since GPT-2, which came out in 2019. While the model won't be open-source, it will be publicly available for anyone to use, modify, or deploy, which is a shift from GPT-3 and GPT-4, which were closed models.

Altman also said that OpenAI hosts developer events to get feedback and lets people experiment with early versions of the model. The first event will occur in San Francisco in the coming weeks, with additional sessions planned in Europe and the Asia-Pacific region.

The company hopes developers will create interesting projects using the model and expects large organizations and governments to run it on their systems. OpenAI is preparing for the model to be modified after its release.

OpenAI’s last “open” model was GPT-2, which was partially released in February 2019, with a full version available later that year. Altman also noted that OpenAI plans to release GPT-4.5 and GPT-5 soon.

The competition in the AI industry is heating up. Rival companies like DeepSeek have launched models similar to ChatGPT, reportedly at a fraction of the cost and time. On March 26, Alibaba unveiled its open-source AI model for cost-effective agents, while Google introduced Gemini 2.5 on March 25. Meta’s Llama AI model, released in February 2023, reached 1 billion downloads, according to CEO Mark Zuckerberg's March 19 post.