OpenAI launched two state-of-the-art open-weight language models, gpt-oss-120b and gpt-oss-20b, that can run locally on devices with just 16 GB of RAM. The 'weights' or internal parameters of open-weight models can be accessed by anyone to know how they process information. The models cannot generate images or videos but can route users' requests to OpenAI's other more powerful closed models.