granite vision 3.3 2b needs about 2.7 GB of memory at Q4_K_M (a 1.5 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 5 quants, from 1.5 GB (Q4_K_M) to 5.1 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 2.7 GB.
Granite Vision 3.3 2B – lightweight vision‑language model for document understanding
Granite Vision 3.3 2B is a compact vision‑language model with 2.53 billion parameters. It is fine‑tuned for visual document tasks such as table and chart analysis, OCR, and general image understanding. The model supports English instructions and images in png or jpeg format.
Good at
- Visual document extraction
- OCR and table parsing
- General image question answering
Limits
- No vision support beyond png/jpeg
- Limited to English instructions
- Requires 8 k context for estimates
Best for
- Enterprise document processing
- Automated data extraction from PDFs
- Image‑based QA in business workflows
Who it’s for. Developers building enterprise applications that need to process visual and textual data. The model is suitable for teams that require efficient, open‑source multimodal capabilities under an Apache 2.0 license.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 1.5 GB | 2.7 GB | Fits | Get file → | |
| Q5_K_M | 1.8 GB | 3.0 GB | Fits | Get file → | |
| Q6_K | 2.1 GB | 3.3 GB | Fits | Get file → | |
| Q8_0 | 2.7 GB | 3.9 GB | Fits | Get file → | |
| BF16 | 5.1 GB | 6.2 GB | Fits | Get file → |
For 8 GB VRAM, Q4_K_M (1.55 GB) uses about 2.7 GB memory. Q5_K_M (1.8 GB) uses about 3 GB. For 12 GB VRAM, Q6_K (2.08 GB) uses about 3.3 GB. Higher precision BF16 (5.07 GB) uses about 6.2 GB.
What GPU do you need for granite vision 3.3 2b?
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 (2.7 GB file, about 3.9 GB in use).
- 6 GB VRAM: the best quant that fits is Q8_0 (2.7 GB file, about 3.9 GB in use).
- 8 GB VRAM: the best quant that fits is BF16 (5.1 GB file, about 6.2 GB in use).
Which granite vision 3.3 2b quant should you choose?
Q4_K_M is the usual starting point for granite vision 3.3 2b: it balances file size and quality well. Here it is a 1.5 GB file.
The Q8_0 file is 2.7 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.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about granite vision 3.3 2b
How much VRAM does granite vision 3.3 2b need?
About 2.7 GB at Q4_K_M with an 8K context window. That is the 1.5 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 granite vision 3.3 2b on an 8 GB GPU?
Yes, with BF16: the 5.1 GB file needs about 6.2 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of granite vision 3.3 2b GGUF?
It depends on the quant: Q4_K_M is 1.5 GB, Q5_K_M is 1.8 GB, Q6_K is 2.1 GB, Q8_0 is 2.7 GB, BF16 is 5.1 GB.
What context length does granite vision 3.3 2b support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.7 GB at 8K and 2.7 GB at 32K.
Where can I download granite vision 3.3 2b GGUF?
The files are on Hugging Face in the ibm-granite/granite-vision-3.3-2b-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 granite vision 3.3 2b 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 Granite Vision 3.3 2B on 8 GB VRAM?
Yes, the Q4_K_M quantization is 1.55 GB and uses about 2.7 GB memory at 8 k context.
Does the model support image segmentation?
Yes, image segmentation is an experimental capability in this model.
What license applies to this model?
The model is released under the Apache 2.0 license.