GPT OSS 120B (Together AI)

GPT OSS 120B is OpenAI's open-weight large language model built around a Mixture-of-Experts architecture with roughly 117 billion total parameters, activating about 5.1 billion per forward pass and tuned to run efficiently on a single high-end GPU using native MXFP4 quantization. It was released alongside a smaller sibling, gpt-oss-20b, as part of OpenAI's effort to bring frontier-style reasoning into openly licensed weights distributed under Apache 2.0. The model is positioned for high-reasoning, agentic, and general-purpose production scenarios, offering configurable reasoning depth, full chain-of-thought access, and native tool use that includes function calling, browsing, and structured output generation.

In practical terms, GPT OSS 120B is well suited for teams that want strong reasoning and agentic capabilities without depending on a closed API, including pipelines that route through external providers such as Eden AI for orchestrated access. The broad 131K token context window supports long-document reasoning and multi-step workflows, while text-in, text-out operation keeps integration straightforward. Its combination of open weights, strong benchmark results relative to similarly sized open models, and emphasis on tool-calling and structured outputs makes it a flexible foundation for building assistants, automated research agents, and developer-facing copilots that need transparency, self-hosting options, and deep reasoning in a single deployable model.