Safety GPT OSS 20B is a 20-billion-parameter open-weights safety and moderation model belonging to OpenAI's gpt-oss family, positioned as a purpose-built classifier rather than a general-purpose chat model. Aggregator listings describe it as a safety model intended for policy screening, content moderation, and risk-aware routing workflows, with text-in and text-out modalities and reasoning support that allows it to evaluate prompts against nuanced policy criteria. This specialization makes it a natural fit as a front-line filter or guardrail component in larger pipelines, where it can triage user inputs and assistant outputs before they reach a downstream language model.

Within the gpt-oss family, Safety GPT OSS 20B slots in alongside the broader 20B and 120B siblings, sharing their large 131,072-token context window and 65,536-token output ceiling while narrowing its role to classification and safety tasks. It exposes reasoning, tool use, structured output, and temperature control, enabling policy decisions that can be returned as parseable JSON for automated moderation systems. Because the weights are open and the model is available across multiple cloud and gateway providers, teams can self-host it for data-sensitive environments or call it as a managed endpoint, using the same model surface for offline policy evaluation and inline runtime screening.