Open-weight guardrail models that run on your infrastructure with 50-100ms latency. No external API calls required.
Lightweight. Edge deployment.
Balanced. General purpose.
High accuracy. Maximum detection.