MOSS Image SFT needs about 4.2 GB of memory at Q4_K (a 3.3 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 7 quants, from 2.2 GB (Q2_K) to 5.7 GB (Q8_0).
MOSS Image SFT – a 5.08B text‑to‑image model in the MOSS family
MOSS Image SFT is a supervised fine‑tuned text‑to‑image model derived from MOSS‑Image‑Base. It supports natural‑language prompts, tag sequences, and JSON layouts for precise text placement and object arrangement. The model uses a FLUX.2 small decoder VAE and runs on NVIDIA GPUs with CUDA.
Good at
- Clear text grouping
- Organized whiteboard layouts
- JSON layout control
Limits
- No vision support
- Requires NVIDIA GPU with CUDA
- License is Apache‑2.0
Best for
- Generating instructional diagrams
- Creating structured whiteboard images
- Producing layout‑driven artwork
Who it’s for. Developers and designers who need precise control over text and object placement in generated images. It is suitable for projects that require structured layouts and clear text rendering.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q2_K | 2.2 GB | 3.0 GB | Fits | Get file → | |
| Q3_K | 2.7 GB | 3.5 GB | Fits | Get file → | |
| Q4_0 | 3.3 GB | 4.2 GB | Fits | Get file → | |
| Q4_K | 3.3 GB | 4.2 GB | Fits | Get file → | |
| Q5_K | 3.9 GB | 4.9 GB | Fits | Get file → | |
| Q6_K | 4.6 GB | 5.6 GB | Fits | Get file → | |
| Q8_0 | 5.7 GB | 6.9 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_0 or Q4_K (file size 3.33 GB, estimated memory 4.2 GB). For 12 GB VRAM, Q5_K (3.92 GB, 4.9 GB) or Q6_K (4.55 GB, 5.6 GB) are suitable. Larger VRAM allows Q8_0 (5.71 GB, 6.9 GB).
What GPU do you need for MOSS Image SFT?
These figures add the file size, an fp16 KV cache at an 8K context window 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 Q3_K (2.7 GB file, about 3.5 GB in use).
- 6 GB VRAM: the best quant that fits is Q6_K (4.6 GB file, about 5.6 GB in use).
- 8 GB VRAM: the best quant that fits is Q8_0 (5.7 GB file, about 6.9 GB in use).
Which MOSS Image SFT quant should you choose?
Q4_K (3.3 GB) is the middle option in this repository. The usual sweet spot between file size and quality, and the quant most people start with.
If memory is tight, the smallest file is Q2_K at 2.2 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 MOSS Image SFT
How much VRAM does MOSS Image SFT need?
About 4.2 GB at Q4_K with an 8K context window. That is the 3.3 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 MOSS Image SFT on an 8 GB GPU?
Yes, with Q8_0: the 5.7 GB file needs about 6.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of MOSS Image SFT GGUF?
It depends on the quant: Q2_K is 2.2 GB, Q3_K is 2.7 GB, Q4_0 is 3.3 GB, Q4_K is 3.3 GB, Q5_K is 3.9 GB, Q6_K is 4.6 GB and 1 more.
Where can I download MOSS Image SFT GGUF?
The files are on Hugging Face in the leejet/MOSS-Image-SFT-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 MOSS Image SFT 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 MOSS Image SFT on 8 GB VRAM?
Yes, using the Q4_0 or Q4_K quantization, which requires about 4.2 GB of VRAM at 8 k context.
Does the model support image input?
No, MOSS Image SFT is a text‑to‑image model and does not accept image inputs.
What license applies to the quantized files?
The quantized files are licensed under Apache‑2.0, matching the original model license.