gpt-oss-120b Overview

gpt-oss-120b is OpenAI's flagship open-weight language model, designed for strong reasoning, coding, and agent-style text workflows outside OpenAI-hosted infrastructure. It uses a sparse mixture-of-experts architecture with 117B total parameters, 5.1B active parameters per token, a native 128K context window, adjustable reasoning effort, support for structured outputs, and strong performance in tool use, function calling, and local or self-hosted deployment setups.