OpenAI has announced plans to launch an open-source artificial intelligence (AI) model, marking the first such release focusing on reasoning capabilities for the San Francisco-based company. The official rollout is anticipated within the coming months. This will be OpenAI’s first open-source model since the introduction of GPT-2 in November 2019. The organization is currently engaging with the developer community to gather input on their requirements and preferences, with an emphasis on safety during the model’s development.
OpenAI Planning to Release an Open-Source AI Model
The landscape of open-source AI has become increasingly competitive, with various players, including Meta, Mistral, Alibaba, and major tech firms like Google and Microsoft, all releasing multiple open models. In contrast, OpenAI has largely refrained from participating in this arena since the launch of GPT-2, opting instead to develop closed models that cannot be downloaded or modified for research or commercial use.
Earlier this year, during an AMA session on Reddit, OpenAI CEO Sam Altman addressed the topic, acknowledging that the company has been “on the wrong side of history” regarding open-source efforts. He indicated the need to devise a new strategy for open-source models, although he noted it was not the company’s highest priority at the time.
On Monday, Altman shared a “Open model feedback” form via a post on X (formerly Twitter), indicating a shift towards developing an open-weight AI model designed for advanced reasoning. The term “weights” refers to the numerical values that represent the connections within the AI, guiding its output generation.
While weights are crucial to the AI model’s functionality, the absence of publicly available datasets and training techniques means that replicating or building upon the model independently remains a challenge. Models that operate in this manner are often regarded as partially open.
Altman emphasized that safety would be a central focus for the model, especially since modifications will occur post-release. OpenAI plans to utilize its Preparedness Framework as a guideline for these safety considerations.
