GPT OSS 120B is a large-scale language model engineered to deliver robust reasoning and versatile performance for general-purpose applications. With 117 billion parameters and 5.1 billion active parameters, the architecture is optimized to provide high-quality outputs while remaining efficient enough to run on a single 80GB GPU, such as the NVIDIA H100 or AMD MI300X. The model is specifically designed for agentic workflows, offering native support for function calling and tool use. Users can also adjust the reasoning effort—ranging from low to high—to balance performance requirements with latency needs, making it a adaptable tool for complex problem-solving.

The model was developed using a combination of reinforcement learning and techniques derived from advanced internal systems, ensuring strong performance on core reasoning benchmarks. It requires the use of the harmony response format to function correctly and provides full access to its chain-of-thought process, which aids in debugging and transparency. Released under a permissive Apache 2.0 license, the model supports fine-tuning and commercial deployment, allowing organizations to customize it for specific needs. Its ability to handle agentic tasks effectively makes it a strong candidate for developers looking to integrate reliable, transparent, and high-reasoning capabilities into their own applications.