SVILA 1 12B needs about 11.6 GB of memory at Q4_K_M (a 7.9 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 11 quants, from 5.1 GB (Q2_K) to 13.6 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 12.9 GB.
SVILA 1 12B – Slovenian Vision Language Model
SVILA 1 12B is a 12.8 billion‑parameter vision‑language model based on gemma3. It was fine‑tuned on a curated Slovenian text‑image dataset. The model supports vision tasks and runs in GGUF format.
Good at
- Vision‑language tasks in Slovenian
- Instruction‑following
- Multimodal generation
Limits
- No vision support beyond image input
- Large context requires 131072 tokens
- MIT license only
Best for
- Local inference on Slovenian datasets
- Multimodal chat assistants
- Image captioning in Slovenian
Who it’s for. Developers needing a Slovenian‑focused vision‑language model for local deployment. The model is available in GGUF format and can be run with llama.cpp on compatible hardware.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q2_K | 5.1 GB | 8.8 GB | Too big | Get file → | |
| Q3_K_S | 5.9 GB | 9.6 GB | Too big | Get file → | |
| Q3_K_M | 6.4 GB | 10.2 GB | Too big | Get file → | |
| Q3_K_L | 6.9 GB | 10.6 GB | Too big | Get file → | |
| IQ4_XS | 7.1 GB | 10.9 GB | Too big | Get file → | |
| Q4_K_S | 7.5 GB | 11.2 GB | Too big | Get file → | |
| Q4_K_M | 7.9 GB | 11.6 GB | Too big | Get file → | |
| Q5_K_S | 8.9 GB | 12.6 GB | Too big | Get file → | |
| Q5_K_M | 9.1 GB | 12.9 GB | Too big | Get file → | |
| Q6_K | 10.5 GB | 14.2 GB | Too big | Get file → | |
| Q8_0 | 13.6 GB | 17.3 GB | Too big | Get file → |
For 8 GB VRAM, use Q2_K (5.1 GB file, 8.8 GB memory) or Q3_K_S (5.89 GB file, 9.6 GB memory). For 12 GB VRAM, Q4_K_M (7.87 GB file, 11.6 GB memory) or Q5_K_S (8.92 GB file, 12.6 GB memory) are suitable. For 16 GB VRAM, Q6_K (10.49 GB file, 14.2 GB memory) or Q8_0 (13.58 GB file, 17.3 GB memory) provide higher quality.
What GPU do you need for SVILA 1 12B?
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: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: the best quant that fits is Q4_K_M (7.9 GB file, about 11.6 GB in use).
- 16 GB VRAM: the best quant that fits is Q6_K (10.5 GB file, about 14.2 GB in use).
- 24 GB VRAM: the best quant that fits is Q8_0 (13.6 GB file, about 17.3 GB in use).
Which SVILA 1 12B quant should you choose?
Q4_K_M is the usual starting point for SVILA 1 12B: it balances file size and quality well. Here it is a 7.9 GB file.
The Q8_0 file is 13.6 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 Q2_K at 5.1 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about SVILA 1 12B
How much VRAM does SVILA 1 12B need?
About 11.6 GB at Q4_K_M with an 8K context window. That is the 7.9 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 SVILA 1 12B on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 8.8 GB, so you would need a larger GPU or CPU offloading.
What is the file size of SVILA 1 12B GGUF?
It depends on the quant: Q2_K is 5.1 GB, Q3_K_S is 5.9 GB, Q3_K_M is 6.4 GB, Q3_K_L is 6.9 GB, IQ4_XS is 7.1 GB, Q4_K_S is 7.5 GB and 5 more.
What context length does SVILA 1 12B support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 3.2 GB at 8K and 12.9 GB at 32K.
Where can I download SVILA 1 12B GGUF?
The files are on Hugging Face in the mradermacher/SVILA-1-12B-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 SVILA 1 12B use?
Hugging Face lists the license as mit. 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 SVILA 1 12B on 8 GB VRAM?
Yes, using Q2_K or Q3_K_S quantizations will fit within 8 GB VRAM with approximate memory usage of 8.8 GB or 9.6 GB respectively.
Does the model support image input?
Yes, SVILA 1 12B has vision capabilities and can process image inputs.
What license does the model use?
The model is released under the MIT license.