OpenAI CEO Sam Altman recently announced on social media that the company will release an open source artificial intelligence model in the next few months. The news has attracted widespread attention, especially after the breakthrough success of the R1 model of Chinese company DeepSeek, OpenAI faces competitive pressure from its peers, especially Meta's Llama series models.
Altman said OpenAI has been considering launching an open source model and the time is now ripe. The open source model means that users can download, modify and run on local hardware for free, rather than being limited to access through cloud services. This approach not only reduces the cost of use, but also provides users with greater flexibility, especially when processing sensitive data.
According to the developers of the DeepSeek model, the training cost of the model is much lower than that of many large AI models, which makes OpenAI realize that its competitiveness in the open source model field needs to be strengthened. Clement Delangue, co-founder and CEO of HuggingFace, welcomed this, believing that the move will make more people aware of the huge potential of open source technology.
Steven Heidel, a member of OpenAI's technical team, also confirmed the plan on social media, saying it will launch models that can run on users' local hardware this year. Meanwhile, OpenAI promises to conduct rigorous testing of the upcoming open source model to ensure it is not used for malicious purposes. Researchers expressed concerns that open source models may be misused in cyber attacks or the development of biological and chemical weapons.
In order to facilitate developers' participation, OpenAI has released a special web page, inviting developers to apply to experience the model in advance. Altman also mentioned that the company will hold developer events related to new model prototypes in the coming weeks to facilitate technology exchanges.
Feedback portal: https://openai.com/open-model-feedback/