Nanonets-OCR-s needs about 2.7 GB of memory at Q4_K_M (a 1.9 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 31 quants, from 0.8 GB (TQ1_0) to 6.2 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.2 GB.
Nanonets-OCR-s: a 3.09B vision‑language model in GGUF format
Nanonets-OCR-s is a 3.09B parameter model based on Qwen2.5‑VL‑3B‑Instruct. It supports vision tasks such as OCR, table extraction, and structured output generation. The model is available in multiple GGUF quantizations for different memory budgets.
Good at
- OCR and text extraction from images
- Structured data extraction from forms and invoices
- Vision‑language reasoning
Limits
- No vision support in quantizations below TQ1_0
- Requires at least 1.6 GB VRAM for TQ1_0 at 8k context
- License information not provided
Best for
- OCR in low‑resource environments
- Extracting tables from scanned documents
- Generating structured JSON from image content
Who it’s for. Developers building OCR or document‑analysis tools who need a lightweight, vision‑enabled model. The model is suitable for edge deployments where GPU memory is limited.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| TQ1_0 | 0.8 GB | 1.6 GB | Fits | Get file → | |
| IQ1_S | 0.9 GB | 1.7 GB | Fits | Get file → | |
| TQ2_0 | 1.0 GB | 1.8 GB | Fits | Get file → | |
| IQ1_M | 1.0 GB | 1.8 GB | Fits | Get file → | |
| IQ2_XXS | 1.0 GB | 1.8 GB | Fits | Get file → | |
| IQ2_XS | 1.1 GB | 1.9 GB | Fits | Get file → | |
| IQ2_S | 1.1 GB | 1.9 GB | Fits | Get file → | |
| IQ2_M | 1.2 GB | 2.0 GB | Fits | Get file → | |
| Q2_K_S | 1.2 GB | 2.0 GB | Fits | Get file → | |
| Q2_K_M | 1.3 GB | 2.1 GB | Fits | Get file → | |
| IQ3_XXS | 1.3 GB | 2.1 GB | Fits | Get file → | |
| IQ3_XS | 1.4 GB | 2.2 GB | Fits | Get file → | |
| IQ3_S | 1.5 GB | 2.3 GB | Fits | Get file → | |
| IQ3_M | 1.5 GB | 2.3 GB | Fits | Get file → | |
| Q3_K_S | 1.5 GB | 2.3 GB | Fits | Get file → | |
| Q3_K_M | 1.6 GB | 2.4 GB | Fits | Get file → | |
| IQ4_XS | 1.7 GB | 2.5 GB | Fits | Get file → | |
| Q4_0 | 1.7 GB | 2.5 GB | Fits | Get file → | |
| IQ4_NL | 1.8 GB | 2.6 GB | Fits | Get file → | |
| Q4_K_S | 1.9 GB | 2.7 GB | Fits | Get file → | |
| Q4_K_M | 1.9 GB | 2.7 GB | Fits | Get file → | |
| Q4_1 | 1.9 GB | 2.7 GB | Fits | Get file → | |
| Q5_0 | 2.1 GB | 2.9 GB | Fits | Get file → | |
| Q5_K_S | 2.2 GB | 3.0 GB | Fits | Get file → | |
| Q5_K_M | 2.2 GB | 3.0 GB | Fits | Get file → | |
| Q5_1 | 2.3 GB | 3.1 GB | Fits | Get file → | |
| Q6_K_M | 2.5 GB | 3.3 GB | Fits | Get file → | |
| Q8_0 | 3.3 GB | 4.1 GB | Fits | Get file → | |
| BF16 | 4.5 GB | 5.3 GB | Fits | Get file → | |
| F16 | 4.5 GB | 5.3 GB | Fits | Get file → | |
| BF16 | 6.2 GB | 7.0 GB | Fits | Get file → |
For 8 GB VRAM, TQ1_0 (0.85 GB) or IQ1_S (0.93 GB) are suitable, estimating 1.6–1.7 GB memory at 8k context. If 12 GB VRAM is available, Q4_K_M (1.93 GB) or Q5_K_M (2.23 GB) provide higher precision with 2.7–3 GB memory usage. Larger models like Q8_0 (3.29 GB) require 4.1 GB memory at 8k context.
What GPU do you need for Nanonets-OCR-s?
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 Q6_K_M (2.5 GB file, about 3.3 GB in use).
- 6 GB VRAM: the best quant that fits is BF16 (4.5 GB file, about 5.3 GB in use).
- 8 GB VRAM: the best quant that fits is BF16 (6.2 GB file, about 7.0 GB in use).
Which Nanonets-OCR-s quant should you choose?
Q4_K_M is the usual starting point for Nanonets-OCR-s: it balances file size and quality well. Here it is a 1.9 GB file.
The Q8_0 file is 3.3 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 TQ1_0 at 0.8 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Nanonets-OCR-s
How much VRAM does Nanonets-OCR-s need?
About 2.7 GB at Q4_K_M with an 8K context window. That is the 1.9 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 Nanonets-OCR-s on an 8 GB GPU?
Yes, with BF16: the 6.2 GB file needs about 7.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Nanonets-OCR-s GGUF?
It depends on the quant: TQ1_0 is 0.8 GB, IQ1_S is 0.9 GB, TQ2_0 is 1.0 GB, IQ1_M is 1.0 GB, IQ2_XXS is 1.0 GB, IQ2_XS is 1.1 GB and 25 more.
What context length does Nanonets-OCR-s support?
The GGUF header lists a trained context length of 128,000 tokens (125K). The KV cache grows with the context you use: about 0.3 GB at 8K and 1.2 GB at 32K.
Where can I download Nanonets-OCR-s GGUF?
The files are on Hugging Face in the Mungert/Nanonets-OCR-s-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
Can I run Nanonets‑OCR‑s on 8 GB VRAM?
Yes. The TQ1_0 and IQ1_S quantizations fit within 8 GB VRAM, using approximately 1.6–1.7 GB at 8k context.
What is the smallest quantization available?
The smallest is TQ1_0, 0.85 GB file size, estimating 1.6 GB memory at 8k context.
Does the model support vision tasks?
Yes, it is a vision‑language model capable of OCR, table extraction, and structured output generation.