«In the coming months, OpenAI plans to release an open source large language model of artificial intelligence — the first since GPT-2. This is stated on a special page on the company’s website, where there is also a form that “developers, researchers, and the entire community” are invited to fill out.
Image source: Growtika / unsplash.com
«“We are excited to collaborate with developers, researchers, and the community to gather feedback and make this model as useful as possible. If you are interested in providing feedback to the OpenAI team, please let us know [via the form] below,” OpenAI says on its website. The company will also gather feedback and show off prototypes of the model at events it will host, the first in San Francisco in a few weeks, followed by events in Europe and Asia-Pacific.
OpenAI has to increasingly fight off attacks from competitors, including the Chinese DeepSeek, which release open AI models. Competitors allow the community to use these systems both for experiments and for commercial purposes. Meta
has invested heavily in the development of Llama family models — in March, these models gained more than 1 billion downloads. DeepSeek quickly gathered a large user base.
«[“I personally think we need to have a different strategy for open source. This is not a view shared by everyone at OpenAI, and it is not our priority right now. [In the future] we will produce better models, but our leadership will be smaller than in previous years,” OpenAI CEO Sam Altman said earlier.
The new open model will support the reasoning function, he added the day before on the X social network. The company will conduct all the standard checks, as before the release of commercial models, and a number of additional ones, given that after the release, users will begin to refine it themselves. It will be deployed by large companies and government agencies, Mr. Altman believes.