Shieldstral
Shieldstral is a 3 billion parameter open weights multimodal safety classifier from Mistral AI. It evaluates text and image content against user supplied, plain language policies at inference time, framing moderation as a binary question answering task and returning calibrated yes or no probability scores without requiring retraining for new policies.
- Type
- Structured data
- Weights
- Closed
- Downloads
- 12k
- Likes
- 278
- License
- apache-2.0
- Library
- vllm
Model hub
mistralai/Shieldstral-1.0-3B
- Repo
- mistralai/Shieldstral-1.0-3B
- Access
- Public
Technical tags and languages
Papers
1
More from Mistral AI
10
-
Codestral 25.01
Text
Codestral 25.01 is Mistral’s code-specialized model, tuned for high-quality generation, completion (including…
-
Codestral Mamba
Text
Codestral Mamba is Mistral AI's 7.3B open-weight code model built on the Mamba2 state-space architecture inst…
-
Devstral 2
Text
Devstral 2 is Mistral AI’s 123B open-weight coding model family, paired with the open-source Mistral Vibe CLI…
-
Devstral Small (24B)
Text
AI model by Mistral AI
-
190 DL
Leanstral 1.5 119B A6B
Coding
Leanstral 1.5 119B A6B is a mixture of experts model with 119 billion total parameters and 6.5 billion active…
-
Magistral Medium
Text
Magistral Medium is Mistral’s mid-tier model in the Magistral line, designed to balance stronger reasoning an…
-
Magistral Medium 1.2
Text
Magistral Medium 1.2 is Mistral AI’s mid-tier reasoning model, designed to balance capability and efficiency.…
-
74k DL
Magistral Small
Text
Magistral Small is a compact model from Mistral’s specialized lineup, tuned for efficiency while still offeri…
About Mistral AI
1
Mistral AI develops frontier generative AI models, infrastructure, and enterprise deployment products.