Winnow 12B needs about 9.7 GB of memory at its main file (a 8.2 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 4 quants, from 0.9 GB (BF16) to 23.8 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.
Winnow 12B – fine‑tuned Gemma 4 12B for typed decisions and image input
Winnow 12B is an EldanRing fine‑tune of Gemma 4 12B IT. It supports typed decisions, chat, and image input without separate adapters. The model is encoder‑free and runs locally via GGUF files.
Good at
- Typed decision scoring
- Image input handling
- Local inference without adapters
Limits
- No vision support beyond images
- Requires 8K context for estimates
- License is Apache‑2.0
Best for
- Decision‑making applications
- Chatbots needing image context
- Local inference on laptops
Who it’s for. Developers building local decision systems or chat interfaces that need image understanding. The model is suitable for environments where a lightweight, fully local solution is required.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| BF16 | 0.9 GB | 2.4 GB | Fits | Get file → | |
| — | 8.2 GB | 9.7 GB | Too big | Get file → | |
| Q8_0 | 12.7 GB | 14.2 GB | Too big | Get file → | |
| BF16 | 23.8 GB | 25.4 GB | Too big | Get file → |
For 8 GB VRAM, use the Q8_0 GGUF (file size 12.67 GB, estimated 14.2 GB memory at 8K context). If 16 GB VRAM is available, the BF16 GGUF (file size 0.86 GB, estimated 2.4 GB memory) is recommended. Larger VRAM allows the 23.83 GB BF16 variant for higher precision.
What GPU do you need for Winnow 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: the best quant that fits is BF16 (0.9 GB file, about 2.4 GB in use).
- 6 GB VRAM: the best quant that fits is BF16 (0.9 GB file, about 2.4 GB in use).
- 8 GB VRAM: the best quant that fits is BF16 (0.9 GB file, about 2.4 GB in use).
- 12 GB VRAM: the best quant that fits is the main file (8.2 GB file, about 9.7 GB in use).
- 16 GB VRAM: the best quant that fits is Q8_0 (12.7 GB file, about 14.2 GB in use).
- 24 GB VRAM: the best quant that fits is Q8_0 (12.7 GB file, about 14.2 GB in use).
- 32 GB VRAM: the best quant that fits is BF16 (23.8 GB file, about 25.4 GB in use).
Which Winnow 12B quant should you choose?
If memory is tight, the smallest file is BF16 at 0.9 GB, with the quality trade-offs that come with fewer bits. The unquantized model. It gives original quality and needs the most memory, so it is mainly for conversion or reference.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Winnow 12B
How much VRAM does Winnow 12B need?
About 9.7 GB at its main file with an 8K context window. That is the 8.2 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 Winnow 12B on an 8 GB GPU?
Yes, with BF16: the 0.9 GB file needs about 2.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Winnow 12B GGUF?
It depends on the quant: BF16 is 0.9 GB, file is 8.2 GB, Q8_0 is 12.7 GB, BF16 is 23.8 GB.
What context length does Winnow 12B support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.
Where can I download Winnow 12B GGUF?
The files are on Hugging Face in the EldanRing/Winnow-12B repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Winnow 12B 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 Winnow 12B on 8 GB VRAM?
Yes, using the Q8_0 quantized GGUF will fit within 8 GB VRAM with the estimated 14.2 GB memory usage at 8K context.
Does Winnow 12B support video input?
No, the model supports image input only; video is not supported.
What license applies to Winnow 12B?
The model is released under the Apache‑2.0 license.