A 21B parameter (3.6B active) open-weight model from OpenAI that targets strong reasoning in a form factor, optimized for low-latency and specialized use cases. It is well suited for mid-scale production workloads that need solid analytical capability, code understanding, and structured writing without the operational overhead of very large models. With its balance of quality and resource usage, GPT-OSS-20B works well as a backbone for internal copilots, domain-specific assistants, and data-heavy reporting tools.
