GPT-OSS is OpenAI’s family of open-weight language models, released under the Apache-2.0 license. It comes in two sizes, gpt-oss-120b and gpt-oss-20b, designed for strong reasoning, agentic workflows, and efficient deployment on everything from a single workstation GPU to larger servers. The release has drawn over 20,000 stars on GitHub and the models were updated in October 2026.
GPT-OSS marked OpenAI’s return to open-weight releases. The models use a mixture-of-experts architecture to deliver high capability per unit of compute: the 120b model targets demanding reasoning and agent workloads, while the 20b model runs comfortably on a single 16GB GPU, putting serious local AI within reach of individual developers. Both support configurable reasoning effort, full chain-of-thought access for debugging, and fine-tuning. Weights are free to download from Hugging Face, and the repository provides reference implementations and guides for running them with common inference stacks.
GPT-OSS is completely free. The weights are published under the Apache-2.0 license, so you can download, run, fine-tune, and even use them commercially without paying OpenAI anything.
GPT-OSS is for developers and teams who want frontier-style reasoning models they can run on their own infrastructure: private deployments, fine-tuned internal assistants, agent builders, and researchers who need open weights rather than an API.