Source: Cointelegraph Original: "{title}"

OpenAI will launch an "open-source" version of its language model this year, allowing developers to run the model on their own hardware.

Company CEO Sam Altman tweeted on March 31 on the X platform that the AI giant will release a "new open-weight language model" with reasoning capabilities in the coming months, but first wants to gather feedback on "how to make it most useful."

"We have been planning this for a long time, but other priorities took precedence. Now is the right time to implement it," he added, noting that this will be the first "open-weight" model since GPT-2 in 2019.

"Some decisions are still pending, so we will hold developer events to gather opinions, and later provide early prototypes for experience."

Source: Sam Altman

An open-weight language model refers to a model that is publicly available for anyone to use, download, modify, or deploy. While it is not as open as a fully open-source model, it represents a significant change compared to the completely closed GPT-3 and GPT-4.

Altman revealed that developer events will kick off in the coming weeks at OpenAI's headquarters in San Francisco, California, followed by a series of meetings in Europe and the Asia-Pacific region.

Source: Steven Heidel

"We look forward to witnessing the innovative results from developers and how large enterprises and governments apply the model in their autonomous operating scenarios,"

Altman stated, "Given that the model may be modified after its release, we will invest additional R&D resources."

The first "open-source" model since GPT-2

OpenAI partially released the open-weight language model GPT-2 in February 2019, completing the full release in November of the same year. Altman also revealed on February 12 that the company plans to launch GPT-4.5 and GPT-5 in the coming weeks or months.

With the release of competitor DeepSeek, the AI arms race continues to heat up—this product, similar to ChatGPT, is reportedly developed at a very low cost and time. On March 26, Alibaba Group launched a new open-source model aimed at cost-effective AI agents, while Google released its latest experimental AI model, Gemini 2.5, on March 25.

Meta CEO Mark Zuckerberg posted on Threads on March 19 that the download count for the Llama series AI models, released by the company in February 2023, has surpassed 1 billion.

Related: SpaceX spacecraft, funded by cryptocurrency investors, successfully launches with crewed polar orbit flight.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。