Mistral releases Shieldstral 1.0 3B safety model

The open-weights tool screens text and images while adapting to custom policies.

MistralReported60 sec read

Apple Magic mouse beside keyboard on white surface

Aug 8: Mistral releases Shieldstral 1.0 3B safety model

Photo: Toms · Unsplash

What happened

Mistral AI released Shieldstral 1.0 3B, a safety classifier model. The model features open weights and supports both text and image inputs. It adapts to specific content moderation rules. Mistral claims it matches performance of models seven times its size.

The context

Guardrail models filter harmful prompts before an AI generates responses. Smaller safety classifiers help developers reduce server costs while enforcing custom moderation rules.

Sources

MarkTechPost ↗ via Google News Reported

See Mistral's Vikshy Score → · Mistral's models & pricing →