GPT OSS 120B (Nebius)

GPT OSS 120B is designed for production, general-purpose, and high-reasoning workloads, with an architecture of 117 billion total parameters and 5.1 billion active parameters. This sparse design is intended to fit on a single 80 GB GPU, making it practical for organizations that need substantial reasoning capacity without the infrastructure pattern of the largest proprietary models.

The model was trained with reinforcement learning and techniques informed by OpenAI’s frontier systems, including o3, and OpenAI reports near-parity with o4-mini on core reasoning benchmarks. Its strengths extend to agentic tool use, few-shot function calling, and chain-of-thought reasoning, with adjustable reasoning effort, parameter fine-tuning, and Harmony-format operation. It is best suited to sophisticated agent workflows and developer applications that can support dedicated high-memory hardware and the model’s required response format.