OpenAI has officially introduced two new open-weight reasoning models: gpt-oss-120b and gpt-oss-20b. This strategic move marks a significant shift for the AI giant, traditionally known for its closed-source approach. The models are released under the permissive Apache 2.0 license, allowing for extensive commercial use, modification, and distribution without the restrictive terms often associated with proprietary AI.

Accompanying the release is a specific gpt-oss usage policy, providing clear guidelines for developers and enterprises. The 'oss' designation likely stands for Open Source Software, signaling a new era of accessibility. By offering these models, OpenAI directly addresses the growing demand for high-performance, transparent AI that can be deployed on-premise or customized for specific needs.

This release empowers developers to build sophisticated applications without relying solely on API-based access, potentially reducing costs and increasing data privacy. The availability of both a massive 120b parameter model and a more lightweight 20b variant ensures flexibility for different hardware constraints and use cases, from large-scale enterprise research to edge computing applications. This development is a direct response to the competitive landscape, where open-weight models from other organizations have gained significant traction.