After OpenAI previously revealed that it would release a new open source artificial intelligence language model this year, more information about this model has been revealed.
Sources indicate that this model, led by Aidan Clark, Vice President of Research at OpenAI, is still in its early stages and could be released as early as this summer. It is expected to achieve state-of-the-art computational performance among open-source models and be positioned as an inference model, similar to OpenAI's current O-series models.
At the same time, OpenAI may adopt a more relaxed open source licensing for this model to compete with Chinese open source models such as DeepSeek. In contrast, Meta's Llama and Google's Gemma, although both are provided in an open source form, have cumbersome terms of use restricting actual use or commercial use.
As for the functionality, this model will support pure text input and output, and can be run on most high-end consumer models. You can even choose to enable inference mode to exchange for more accurate inference and answer capabilities, but this will relatively delay execution efficiency.
In addition to the new version of the open source model planned to be released this time, OpenAI also seems to plan to release more open source models, even including smaller model versions, to attract more people to use them.
