OpenAI open-weight models (gpt-oss)
Deployment options and support resources for open-weight models.
In this article
Where models run
gpt-oss-120b and gpt-oss-20b are reasoning models designed to run on infrastructure you control or with a hosting provider. They are not served through the OpenAI API or inside ChatGPT.
Getting started
Obtain the weights and reference resources, select a compatible inference runtime and confirm the hardware requirements. Common open tooling supports self-managed or cloud deployments and customization.
Licensing and support
The weights use the Apache 2.0 license, with applicable usage-policy requirements. Technical guides, reference code and community resources cover deployment; OpenAI API rate limits and pricing do not apply to these self-hosted models.