Mistral AI has released Shieldstral, a new 3-billion parameter open-weights model designed for multimodal content moderation. This model can adapt to various safety policies by accepting plain-language questions at inference time, unifying the evaluation of text and image content without requiring retraining. Shieldstral offers calibrated safety scores and is efficient enough to run on a single 16GB NVIDIA GPU, making it accessible for developers. AI
IMPACT Provides an efficient, adaptable multimodal moderation solution for developers, potentially improving AI safety across various applications.
RANK_REASON Frontier-lab model release with system card
Read on Mastodon — fosstodon.org →
AI-generated summary · Google Gemini · from 5 sources. How we write summaries →