OpenAI has introduced two new artificial intelligence models, gpt-oss-120b and gpt-oss-20b, which are available for free download and customization. This strategic shift aims to enhance its competitive stance against similar offerings from both U.S. and Chinese companies. The announcement was made on October 10, 2023, as OpenAI responds to increasing calls for transparency regarding its technology.
The new models, classified as “open-weight language models,” allow users to access and modify the underlying trained parameters. This initiative aligns with OpenAI’s foundational mission, as articulated by Chief Executive Officer Sam Altman: “Going back to when we started in 2015, OpenAI’s mission is to ensure AGI (Artificial General Intelligence) that benefits all of humanity.”
Open-Weight Models and Market Implications
An open-weight model in generative AI facilitates user-driven modifications, setting it apart from proprietary systems. Meta has championed an open-source approach to AI, while Chinese startup DeepSeek has gained attention for offering a customizable model at a low cost. During a media briefing, OpenAI Co-founder and President Greg Brockman emphasized the significance of this release, stating, “This is the first time that we’re releasing an open-weight model in language in a long time, and it’s really incredible.”
The text-only models are designed for various applications, including internet searching and executing computer code, and are optimized for efficient operation on local systems. OpenAI expressed optimism that these new tools will spur innovative research and product development. Altman noted, “We are quite hopeful that this release will enable new kinds of research and the creation of new kinds of products.”
Collaboration and Ethical Considerations
OpenAI is collaborating with partners such as Orange, a French telecommunications giant, and Snowflake, a cloud-based data platform, to explore practical applications for the new models. To mitigate potential misuse, the open-weight models have been specifically tuned to prevent malicious applications.
Earlier this year, Altman acknowledged that OpenAI had been “on the wrong side of history” regarding transparency in its technology. In response to internal and external pressures, he reaffirmed that OpenAI would remain a nonprofit organization, abandoning previous plans to transition to a for-profit model. This decision followed significant criticism from AI safety advocates and former co-founder Elon Musk, who has publicly voiced concerns over the direction of the company he exited in 2018.
The current structure allows OpenAI’s for-profit arm to generate revenue while remaining under the oversight of a nonprofit board. This revised approach aims to balance the need for financial sustainability with the commitment to its original mission of advancing AI for the benefit of humanity.
As OpenAI navigates these changes, the release of its open-weight models marks a pivotal moment in the AI landscape, promising to influence both research and product development significantly.
