Sansar Ocr 2b needs about 2.2 GB of memory at Q4_K_M (a 1.3 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 12 quants, from 1.0 GB (Q2_K) to 3.8 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.6 GB.
Sansar OCR 2B – a 1.88B‑parameter vision‑language model for Sanskrit OCR
Sansar OCR 2B is a 1.88B‑parameter Qwen3.5‑based model fine‑tuned on 19,377 Sanskrit page images. It reads printed Sanskrit book pages and outputs line‑by‑line text with a median letter error rate of 1.05%. The model supports reasoning, tools, and vision.
Good at
- Sanskrit OCR
- Line‑by‑line text extraction
- Reasoning on scanned pages
Limits
- No commercial AI training labels
- License cc‑by‑nc‑4.0
- No native vision support beyond OCR
Best for
- Digitizing Sanskrit books
- Creating searchable text from scanned pages
- Research in Sanskrit NLP
Who it’s for. Researchers and developers working with Sanskrit texts who need accurate OCR. The model is suitable for academic projects and digital humanities initiatives that require line‑level text extraction from Devanagari scans.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q2_K | 1.0 GB | 1.9 GB | Fits | Get file → | |
| Q3_K_S | 1.0 GB | 1.9 GB | Fits | Get file → | |
| Q3_K_M | 1.1 GB | 2.0 GB | Fits | Get file → | |
| Q3_K_L | 1.2 GB | 2.1 GB | Fits | Get file → | |
| IQ4_XS | 1.2 GB | 2.1 GB | Fits | Get file → | |
| Q4_K_S | 1.2 GB | 2.1 GB | Fits | Get file → | |
| Q4_K_M | 1.3 GB | 2.2 GB | Fits | Get file → | |
| Q5_K_S | 1.4 GB | 2.3 GB | Fits | Get file → | |
| Q5_K_M | 1.4 GB | 2.3 GB | Fits | Get file → | |
| Q6_K | 1.6 GB | 2.5 GB | Fits | Get file → | |
| Q8_0 | 2.0 GB | 2.9 GB | Fits | Get file → | |
| F16 | 3.8 GB | 4.7 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_K_M (1.27 GB, ~2.2 GB memory) or Q3_K_M (1.1 GB, ~2 GB). For 4 GB VRAM, Q3_K_S (1.02 GB, ~1.9 GB) or Q2_K (0.97 GB, ~1.9 GB) are suitable. Higher‑quality F16 (3.78 GB, ~4.7 GB) requires 8 GB+ VRAM.
What GPU do you need for Sansar Ocr 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 (2.0 GB file, about 2.9 GB in use).
- 6 GB VRAM: the best quant that fits is F16 (3.8 GB file, about 4.7 GB in use).
Which Sansar Ocr 2b quant should you choose?
Q4_K_M is the usual starting point for Sansar Ocr 2b: it balances file size and quality well. Here it is a 1.3 GB file.
The Q8_0 file is 2.0 GB, about 1.6 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 Q2_K at 1.0 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 Sansar Ocr 2b
How much VRAM does Sansar Ocr 2b need?
About 2.2 GB at Q4_K_M with an 8K context window. That is the 1.3 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 Sansar Ocr 2b on an 8 GB GPU?
Yes, with F16: the 3.8 GB file needs about 4.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Sansar Ocr 2b GGUF?
It depends on the quant: Q2_K is 1.0 GB, Q3_K_S is 1.0 GB, Q3_K_M is 1.1 GB, Q3_K_L is 1.2 GB, IQ4_XS is 1.2 GB, Q4_K_S is 1.2 GB and 6 more.
What context length does Sansar Ocr 2b support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.4 GB at 8K and 1.6 GB at 32K.
Where can I download Sansar Ocr 2b GGUF?
The files are on Hugging Face in the mradermacher/sansar-ocr-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 Sansar Ocr 2b use?
Hugging Face lists the license as cc-by-nc-4.0. 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 Sansar OCR 2B on 8 GB VRAM?
Yes, using Q4_K_M or Q3_K_M quantizations will fit within 8 GB VRAM with approximate memory usage of 2.2 GB or 2 GB.
Does the model support other languages?
The model is specialized for Sanskrit (Devanagari) and does not support other languages out of the box.