MistralAIAugust 4, 2026

Shieldstral 1.0

Mistral AI’s open-weight multimodal safety classifier, released 2026-08-04; status: in public preview.

PreviewText + Image inputText outputOpen weights

At a glance

Context window
32K tokens
Max output
—
Text input / output · per 1M tokens
— / —
Knowledge cutoff
—

Specifications

Provider
MistralAI
Model type
Multimodal model
Released
August 4, 2026
Total parameters
3.8B
Architecture
Dense
Context window
32K tokens
Input
Text, Image
Output
Text
Weights
Open weights
Licence
Apache 2.0
Commercial use
Commercial use allowed

Model analysis

Shieldstral 1.0 is a compact multimodal moderation model released by Mistral AI in public preview on August 4, 2026, with Apache 2.0 weights on Hugging Face (mistralai/Shieldstral-1.0). It handles prompt moderation, response moderation, prompt-response pair classification, refusal detection and safety filtering, answering natural-language policy questions with yes or no rather than using a fixed taxonomy.

Specifications

3.8B parameters (dense), 32K-token context, text and image input, yes/no output. Mistral lists no API model ID, price, memory requirement or benchmark results; weights are the only deployment option.

Source: Mistral model page.

Last reviewed 2026-10-19

Community discussion

Share the API, setup, and tradeoffs you actually tested. Rate the model once; comment as often as you like.

Used this model? Sign in with Google to rate it and share what worked, what did not, and which API or setup you tested.

No community comments yet.