UI TARS 2B SFT needs about 1.9 GB of memory at Q4_K_M (a 1.1 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 4 quants, from 1.0 GB (Q3_K_L) to 1.9 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 0.9 GB.
UI TARS 2B SFT – a 1.78B vision‑language model for GUI automation
UI TARS 2B SFT is a 1.78B parameter model based on Qwen2VL. It is finetuned for native GUI agent tasks, combining perception, reasoning, grounding, and memory in a single VLM. The model supports vision and operates with a 32,768 token context window.
Good at
- GUI automation
- Vision‑language reasoning
- End‑to‑end task execution
Limits
- No vision support beyond GUI
- Large VRAM requirement for higher quant
- License is apache‑2.0
Best for
- Automated desktop workflows
- GUI testing scripts
- Interactive agent demos
Who it’s for. Developers building autonomous GUI agents and researchers exploring vision‑language interaction. The model is suitable for those who need a single VLM that can perceive and act within graphical interfaces.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q3_K_L | 1.0 GB | 1.7 GB | Fits | Get file → | |
| Q4_K_M | 1.1 GB | 1.9 GB | Fits | Get file → | |
| Q6_K | 1.5 GB | 2.2 GB | Fits | Get file → | |
| Q8_0 | 1.9 GB | 2.6 GB | Fits | Get file → |
For 8 GB VRAM, the Q4_K_M quant (~1.12 GB file) uses about 1.9 GB memory at 8 k context, which is within limits. If more VRAM is available, Q6_K (~1.46 GB) uses about 2.2 GB. Q3_K_L is the smallest but uses 1.7 GB memory.
What GPU do you need for UI TARS 2B SFT?
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 Q8_0 (1.9 GB file, about 2.6 GB in use).
Which UI TARS 2B SFT quant should you choose?
Q4_K_M is the usual starting point for UI TARS 2B SFT: it balances file size and quality well. Here it is a 1.1 GB file.
The Q8_0 file is 1.9 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_L at 1.0 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 UI TARS 2B SFT
How much VRAM does UI TARS 2B SFT need?
About 1.9 GB at Q4_K_M with an 8K context window. That is the 1.1 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 UI TARS 2B SFT on an 8 GB GPU?
Yes, with Q8_0: the 1.9 GB file needs about 2.6 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of UI TARS 2B SFT GGUF?
It depends on the quant: Q3_K_L is 1.0 GB, Q4_K_M is 1.1 GB, Q6_K is 1.5 GB, Q8_0 is 1.9 GB.
What context length does UI TARS 2B SFT support?
The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 0.2 GB at 8K and 0.9 GB at 32K.
Where can I download UI TARS 2B SFT GGUF?
The files are on Hugging Face in the lmstudio-community/UI-TARS-2B-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 UI TARS 2B 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 UI TARS 2B SFT on 8 GB VRAM?
Yes. The Q4_K_M quant uses about 1.9 GB memory at 8 k context, fitting within 8 GB VRAM. The Q3_K_L quant is even smaller, using about 1.7 GB.
What context length does the model support?
The model was trained with a 32,768 token context window.
Is the model licensed for commercial use?
It is released under the apache‑2.0 license, which permits commercial use.