OpenAI on Tuesday released two text-only models, gpt-oss-120b and gpt-oss-20b, as lower-cost options for developers, researchers, companies, and governments.
The open-weight language models, the first from OpenAI since it introduced GPT-2 in 2019, comes after Chinese startup DeepSeek, Meta Platforms Inc., and Mistral AI released similar-type models in recent years. OpenAI had repeatedly delayed the launch so it could “run additional safety tests and review high-risk areas,” company CEO Sam Altman recently said on X.
“It’s been exciting to see an ecosystem develop, and we are excited to contribute to that and really push the frontier and then see what happens from there,” OpenAI President Greg Brockman said in a briefing announcing the latest from OpenAI.
“OpenAI showed the world what could be built on NVIDIA AI — and now they’re advancing innovation in open-source software,” NVIDIA CEO Jensen Huang said in a statement. “The gpt-oss models let developers everywhere build on that state-of-the-art open-source foundation, strengthening U.S. technology leadership in AI — all on the world’s largest AI compute infrastructure.”
The open models can send complex queries to artificial intelligence (AI) models in the cloud if they are unable to perform tasks such as processing an image, OpenAI said. In the cloud, developers can connect the open model to one of the company’s more capable closed models, according to OpenAI.
OpenAI’s new models can be downloaded on platforms Hugging Face and GitHub under an Apache 2.0 license. The models can run on PCs through programs like LM Studio and Ollama. Cloud providers Amazon.com Inc. and Microsoft are also making the models available. Additionally, Cerebras Systems announced inference support for the models.
OpenAI said its new open models are trained using high-compute reinforcement learning, a post-training process to teach AI models right from wrong in simulated environments that use large clusters of NVIDIA GPUs. This will allow them to power AI agents, as well as call on tools like a web search or Python code execution as part of its chain-of-thought process.
The release of gpt-oss is designed to land favorably with developers and the Trump Administration, both of whom watched Chinese AI labs rise to prominence in the open-source space. Last month the Trump Administration urged AI developers in the U.S. to open source more technology to promote global adoption of AI hewing to American values.
“OpenAI releasing their open source models is a clear signal: Enterprises want AI on their servers, under their control,” Jesse Williams, chief operating officer at Jozu, said in an email. “For security-focused enterprises, the future isn’t just about who has the best model—it’s about maintaining the ability to deploy it safely in their own environment, whether that is on-prem or in a private cloud.”
