Mistral AI Releases Shieldstral Safety Model
French AI company announces open-weights multimodal safety classifier for on-device deployment.
Mistral AI posted an announcement introducing Shieldstral, its 3B open-weights model for content safety that runs on-device. The company account described the model as multimodal and stated it achieves leading benchmark scores while operating on a single 16GB NVIDIA GPU. Enterprises gain control over safety criteria through customization. The same details appear in a thread and on the linked company page. Other posts quoted the announcement, including one labeling the project Le Cuck’ Petit internally.
🛡️Introducing Shieldstral, Mistral’s 3B open-weights model for content safety that can be deployed on-device 🧵 http://mistral.ai/news/shieldstral
