Shieldstral — local AI model by Mistral AI
Written by Jakub Rusinowski · Last updated
Mistral's open-weight moderation classifier for text and images. You state your policy as a yes/no question in each request and it returns a calibrated safety score, so one model serves many policies without retraining. Apache-2.0.
Variants
The smallest Shieldstral variant needs about 3 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| Shieldstral 1.0 3B → 3.8B | ~3.1 GB | 32,768 | cloud-hosted tag |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Shieldstral locally
Install Ollama, then pull the tag.
No Ollama tag. Mistral documents vLLM and llama.cpp builds on the model card.
Pick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: Shieldstral 1.0 3BRecommended GPU
The cheapest catalogued GPU that runs Shieldstral locally (min 3 GB VRAM) is the Intel Arc B570 (10 GB).
Shieldstral — frequently asked questions
How much VRAM does Shieldstral need?
Shieldstral needs about 3 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Shieldstral 1.0 3B (3 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Shieldstral on an RTX 4090 (24 GB)?
Yes — Shieldstral runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for Shieldstral?
Q4_K_M is the best balance of quality and VRAM for Shieldstral in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Shieldstral with Ollama?
Shieldstral has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.