OpenAI has announced plans to release its first open language model since GPT-2. In a statement on its website, OpenAI invited developers, researchers, and the broader community to fill out a feedback form to share their wishes regarding the new model. The company plans to hold a series of developer events to gather feedback and showcase model prototypes. The first event will take place in San Francisco in the coming weeks, followed by events in Europe and the Asia-Pacific region.
OpenAI faces increasing pressure from competitors such as the Chinese lab DeepSeek, which adheres to an open approach to model releases. This strategy has proven successful for many companies. For example, Meta announced that its Llama model family has already been downloaded over one billion times. DeepSeek has also quickly attracted a large user base worldwide and caught the attention of investors.
OpenAI CEO Sam Altman recently admitted in a discussion on Reddit that the company needs a different strategy regarding open source. Altman noted that they want to build better models, but will not have the same advantage as in previous years. He also stated that the upcoming open model from OpenAI will have “reasoning” capabilities similar to o3-mini and will be evaluated according to their preparatory system before release.
OpenAI hopes the new model will be useful for developers, large companies, and governments wishing to run it on their own hardware. The company aims to learn what solutions developers can create and how large organizations will use this technology.
