Granite Guardian 3.0 2B

About

IBM Granite Guardian 3.0 2B is the smaller variant of the Granite Guardian 3.0 safety model family. Detects risk dimensions including harm, social bias, jailbreaking, violence, profanity, sexual content, and unethical behavior. Also covers RAG hallucination risks: context relevance, groundedness, and answer relevance. Outputs binary yes/no with probability scores. Optimized for cost/latency-sensitive applications. English only. Apache 2.0.

Granite Guardian 3.0 2B has a 4K-token context window.

Capabilities

Rankings