What it is
GPT-OSS 120B is an open-weight reasoning model from OpenAI. Its importance is not only the model size; it is the option to place more of the inference and deployment boundary under a developer’s control.
Why it matters
Open-weight models change the questions a team can ask. Instead of choosing only between hosted endpoints, teams can consider hardware, serving stack, fine-tuning, inspection and data locality as part of the product decision.
Who should look at it?
Infrastructure teams, model-serving developers and organisations with strict control requirements should start with the official release. Agent builders may also care because local or controlled inference changes how tools, data and latency are designed.
What to verify
Use the official announcement for the license, supported deployment paths, capabilities, hardware expectations and limitations. Do not infer production cost or quality from the model name alone.