Mistral AI Releases Shieldstral 1.0 3B: An Open-Weights Policy-Adaptive Multimodal Safety Classifier Matching Models 7× Its Size

Mistral AI has released Shieldstral 1.0 3B, an open-weights multimodal safety classifier that allows users to define moderation policies using plain-language queries at inference time. By treating content moderation as a flexible yes/no question rather than a rigid taxonomy, the 3-billion-parameter model can match the performance of systems seven times its size. This approach offers developers a more efficient, adaptable way to enforce custom safety standards across multimodal data without requiring extensive model retraining.