OvisOCR2 needs about 1.4 GB of memory at Q4_K_M (a 0.5 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 9 quants, from 0.5 GB (Q3_K_M) to 1.5 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.6 GB.
Compact 0.8B vision‑language model for page‑level OCR
OvisOCR2 is a 0.8B end‑to‑end model that parses document pages into Markdown. It handles text, formulas, tables, and visual regions in natural reading order. The model is released under Apache‑2.0.
Good at
- OCR of complex layouts
- Markdown generation with LaTeX
- Table and formula extraction
Limits
- No vision support without mmproj
- Large VRAM needed for higher quant
- Requires llama‑cpp CLI for multimodal
Best for
- Document digitization
- Academic paper conversion
- Legal document parsing
Who it’s for. Developers needing a lightweight OCR model that outputs structured Markdown. Researchers working on document understanding can use it as a single‑stage alternative to pipeline methods.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q3_K_M | 0.5 GB | 1.4 GB | Fits | Get file → | |
| Q4_K_S | 0.5 GB | 1.4 GB | Fits | Get file → | |
| Q4_K_M | 0.5 GB | 1.4 GB | Fits | Get file → | |
| Q5_K_S | 0.6 GB | 1.5 GB | Fits | Get file → | |
| Q5_K_M | 0.6 GB | 1.5 GB | Fits | Get file → | |
| Q6_K | 0.6 GB | 1.5 GB | Fits | Get file → | |
| Q8_0 | 0.8 GB | 1.7 GB | Fits | Get file → | |
| F16 | 1.5 GB | 2.4 GB | Fits | Get file → | |
| BF16 | 1.5 GB | 2.4 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_K_M or Q5_K_S (est. 1.4–1.5 GB). For 12 GB VRAM, Q6_K or Q8_0 (est. 1.5–1.7 GB) provide higher quality. The unquantized F16/BF16 files need 2.4 GB VRAM and are best for maximum accuracy.
What GPU do you need for OvisOCR2?
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 BF16 (1.5 GB file, about 2.4 GB in use).
Which OvisOCR2 quant should you choose?
Q4_K_M is the usual starting point for OvisOCR2: it balances file size and quality well. Here it is a 0.5 GB file.
The Q8_0 file is 0.8 GB, about 1.5 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_M at 0.5 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 OvisOCR2
How much VRAM does OvisOCR2 need?
About 1.4 GB at Q4_K_M with an 8K context window. That is the 0.5 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 OvisOCR2 on an 8 GB GPU?
Yes, with BF16: the 1.5 GB file needs about 2.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of OvisOCR2 GGUF?
It depends on the quant: Q3_K_M is 0.5 GB, Q4_K_S is 0.5 GB, Q4_K_M is 0.5 GB, Q5_K_S is 0.6 GB, Q5_K_M is 0.6 GB, Q6_K is 0.6 GB and 3 more.
What context length does OvisOCR2 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 OvisOCR2 GGUF?
The files are on Hugging Face in the Abiray/OvisOCR2-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 OvisOCR2 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 OvisOCR2 on 8 GB VRAM?
Yes, using Q4_K_M or Q5_K_S quantizations, which estimate 1.4–1.5 GB VRAM at 8k context.
Do I need a separate image projector?
Yes, OvisOCR2 requires an mmproj file such as mmproj-F16.gguf to process images.