BiRefNet_lite needs about 1.0 GB of memory at F16 (a 0.4 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 3 quants, from 0.1 GB (F16) to 0.4 GB (F16).
BiRefNet_lite: lightweight image segmentation model
BiRefNet_lite is a 0.22B parameter model for dichotomous image segmentation. It is released under the MIT license and is optimized for consumer hardware using vision.cpp. The model is derived from the original BiRefNet architecture.
Good at
- High‑resolution background removal
- Fast inference on low‑VRAM GPUs
- Lightweight deployment
Limits
- No vision support beyond image segmentation
- Requires 8k context for memory estimates
- No mixture of experts
Best for
- Mobile photo editing
- Real‑time background removal in apps
- Batch segmentation of large image sets
Who it’s for. Developers building lightweight image segmentation tools on consumer GPUs. Researchers needing a small, open‑source model for rapid prototyping.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| F16 | 0.1 GB | 0.6 GB | Fits | Get file → | |
| F16 | 0.4 GB | 1.0 GB | Fits | Get file → | |
| F16 | 0.4 GB | 1.0 GB | Fits | Get file → |
For 8 GB VRAM, the 0.09 GB F16 quant uses about 0.6 GB memory at 8k context. The 0.44 GB F16 quant uses about 1 GB memory. Choose the smaller quant for tighter budgets, the larger for higher accuracy.
What GPU do you need for BiRefNet_lite?
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 F16 (0.4 GB file, about 1.0 GB in use).
Which BiRefNet_lite quant should you choose?
F16 (0.4 GB) is the middle option in this repository. The unquantized model. It gives original quality and needs the most memory, so it is mainly for conversion or reference.
If memory is tight, the smallest file is F16 at 0.1 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 BiRefNet_lite
How much VRAM does BiRefNet_lite need?
About 1.0 GB at F16 with an 8K context window. That is the 0.4 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 BiRefNet_lite on an 8 GB GPU?
Yes, with F16: the 0.4 GB file needs about 1.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of BiRefNet_lite GGUF?
It depends on the quant: F16 is 0.1 GB, F16 is 0.4 GB, F16 is 0.4 GB.
Where can I download BiRefNet_lite GGUF?
The files are on Hugging Face in the Acly/BiRefNet-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 BiRefNet_lite 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 BiRefNet_lite on 4 GB VRAM?
The 0.09 GB quant uses 0.6 GB memory at 8k context, so it fits on 4 GB VRAM with headroom.
Is BiRefNet_lite suitable for real‑time inference?
Yes, its small size and F16 quantization enable fast inference on consumer GPUs.