Best Moderation
Moderation models that read a policy and return a verdict on the content you hand them, for text and in some cases images. Useful in front of a generation pipeline, in review queues, and anywhere the rules change per product or per customer instead of following a fixed list of categories.
Best rated
by OpenAI
gpt-oss-safeguard-20b is OpenAI’s open-weight safety classifier. Instead of a fixed taxonomy, it takes the policy in the prompt and labels each item against it, so the same model can serve teams whose rules differ. It reasons before answering, which leaves a trace of why an item was labelled the way it was, and its 8K context window holds a policy alongside the content under review. It fits moderation pipelines and review queues where the rules change per product or per customer.
Featured Models
Top-performing models in this category, recommended by our community and performance benchmarks.
by Mistral AI
Shieldstral 1.0 3B is Mistral AI’s moderation model for text and images. The policy travels in the prompt as plain-language questions, so the same model covers products whose rules differ without retraining or a fixed list of harm categories. At 3B parameters it stays cheap enough to sit in front of every request, and its 8K context window holds the policy alongside the content under review. It fits moderation pipelines, review queues, and screening of user uploads before they reach a generation model.