Data as of Jul 25, 2026 · Based on 2,718,867 AI responses across 9,511 prompts · See how Parse measures this
AprielGuardhuggingface.coAprielGuard is an 8B-parameter safeguard model from ServiceNow-AI on Hugging Face designed to detect and mitigate safety risks (toxicity, bias, misinformation) and security threats (prompt injections, jailbreaks, indirect prompt attacks) in LLM interactions, unifying safety and adversarial risk under a single taxonomy and training framework for standalone prompts, multi-turn conversations, and agentic workflows. It offers structured reasoning traces with reasoning-on and off modes and agentic-aware moderation to identify threats in reasoning/planning chains, tool-use sequences, and API executions, and is lightweight for production deployment. Built on a downscaled Apriel-1.5-15B base model, AprielGuard underwent extensive supervised fine-tuning on over 600,000 high-quality samples, trained on diverse synthetic data, and evaluated on a range of standard safety and adversarial benchmarks.
Parse Score
Sources
arxiv.org shapes more of what AI says about AprielGuard than any other source, at 100% of its citations.
