OpenAI releases gpt-oss-120b and gpt-oss-20b, open-weight models under Apache 2.0
OpenAI's two open-weight reasoning models use the Apache 2.0 licence and can run on a single high-end GPU or on a single workstation.
OpenAI has released gpt-oss-120b and gpt-oss-20b, two open-weight reasoning models under the Apache 2.0 licence. As described by OpenAI, the larger model can run on a single high-end GPU, while the smaller one is light enough to run on a single workstation.
Open-weight means the model’s parameter set is published so that organisations can download it and run it on their own infrastructure, instead of calling the provider’s API. The Apache 2.0 licence permits use and modification with fewer restrictions than a proprietary licence. Two sizes allow an appropriate level of resources to be chosen.
For businesses, the appeal is that data does not have to leave internal infrastructure, which suits organisations with security requirements or an unwillingness to depend on an external service. Being able to run on a single workstation also lowers the barrier to experimentation.
What to keep in mind is that running a model in-house brings infrastructure responsibility with it: GPU configuration, memory, response speed and power costs all fall to the organisation. The announcement gave no performance figures or detailed configuration requirements, so testing should be done on the hardware actually intended for use.
Source: openai.com
