Mistralai Shieldstral 1.0 3B needs about 3.5 GB of memory at Q4_K_M (a 2.1 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 7 quants, from 1.8 GB (Q3_K_M) to 3.7 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 3.5 GB.
Compact 3B policy‑adaptive multimodal safety classifier
Shieldstral 1.0 3B is a 3.43‑billion‑parameter model that evaluates content against a natural‑language safety policy. It supports text‑only, image‑only, and text‑plus‑image moderation in a single forward pass. The model is built on Mistral‑3 architecture and trained on up to 262144 tokens.
Good at
- Policy‑adaptive moderation
- Multimodal text and image classification
- Single‑token continuous safety score
Limits
- Requires a vision projector for image moderation
- Context window best kept within 32k tokens
- No vision support without mmproj
Best for
- Real‑time edge moderation
- Prompt and response filtering
- Model refusal detection
Who it’s for. Developers building lightweight moderation pipelines on consumer GPUs or CPUs. The model is open‑licensed under Apache‑2.0, making it suitable for commercial and non‑commercial use.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q3_K_M | 1.8 GB | 3.2 GB | Fits | Get file → | |
| Q4_K_S | 2.1 GB | 3.4 GB | Fits | Get file → | |
| Q4_K_M | 2.1 GB | 3.5 GB | Fits | Get file → | |
| Q5_K_S | 2.4 GB | 3.8 GB | Fits | Get file → | |
| Q5_K_M | 2.5 GB | 3.8 GB | Fits | Get file → | |
| Q6_K | 2.8 GB | 4.2 GB | Fits | Get file → | |
| Q8_0 | 3.7 GB | 5.0 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_K_M (file size 2.15 GB, estimated 3.5 GB memory at 8k context). If 12 GB VRAM is available, Q5_K_M (2.47 GB, 3.8 GB memory) offers higher quality. Avoid Q8_0 unless you have 16 GB+ VRAM.
What GPU do you need for Mistralai Shieldstral 1.0 3B?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is Q5_K_M (2.5 GB file, about 3.8 GB in use).
- 6 GB VRAM: the best quant that fits is Q8_0 (3.7 GB file, about 5.0 GB in use).
Which Mistralai Shieldstral 1.0 3B quant should you choose?
Q4_K_M is the usual starting point for Mistralai Shieldstral 1.0 3B: it balances file size and quality well. Here it is a 2.1 GB file.
The Q8_0 file is 3.7 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is Q3_K_M at 1.8 GB, with the quality trade-offs that come with fewer bits. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Mistralai Shieldstral 1.0 3B
How much VRAM does Mistralai Shieldstral 1.0 3B need?
About 3.5 GB at Q4_K_M with an 8K context window. That is the 2.1 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Mistralai Shieldstral 1.0 3B on an 8 GB GPU?
Yes, with Q8_0: the 3.7 GB file needs about 5.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Mistralai Shieldstral 1.0 3B GGUF?
It depends on the quant: Q3_K_M is 1.8 GB, Q4_K_S is 2.1 GB, Q4_K_M is 2.1 GB, Q5_K_S is 2.4 GB, Q5_K_M is 2.5 GB, Q6_K is 2.8 GB and 1 more.
What context length does Mistralai Shieldstral 1.0 3B support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.9 GB at 8K and 3.5 GB at 32K.
Where can I download Mistralai Shieldstral 1.0 3B GGUF?
The files are on Hugging Face in the Abiray/Shieldstral-1.0-3B-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Mistralai Shieldstral 1.0 3B use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Shieldstral on 8 GB VRAM?
Yes, using the Q4_K_M quantization which requires about 3.5 GB of VRAM at 8k context.
Do I need a separate vision projector for image moderation?
Yes, a mmproj file is required for multimodal use; the projector is kept in full precision.
What context length should I use?
Keep the context within the trained 262144 tokens, preferably under 32k for best performance.