gpt-oss is a family of open-weight language models from OpenAI, built upon the Mixture-of-Experts (MoE) architecture. This playground allows you to experiment with the gpt-oss-120b and gpt-oss-20b models, explore their capabilities.