MoganOCR-TR

TurkishOCR-Bench Buy Me a Coffee

MoganOCR-TR-GGUF

GGUF files of MoganOCR-TR for llama.cpp. MoganOCR-TR is a 2B-parameter OCR model for Turkish documents: it takes one page image and writes the whole page in reading order as Markdown, with tables as HTML and formulas as LaTeX.

Use one of the model files together with mmproj-MoganOCR-TR-F16.gguf, the vision encoder.

File Size
MoganOCR-TR-Q8_0.gguf (recommended) 2.01 GB
MoganOCR-TR-Q6_K.gguf 1.70 GB
MoganOCR-TR-Q5_K_M.gguf 1.56 GB
MoganOCR-TR-Q4_K_M.gguf 1.44 GB
MoganOCR-TR-BF16.gguf 3.78 GB
mmproj-MoganOCR-TR-F16.gguf 0.67 GB

Usage

llama-server -hf moganai/MoganOCR-TR-GGUF:Q8_0 --image-max-tokens 4096 -c 16384

Send one page image per request with the prompt Sayfayı markdown olarak yaz. and temperature 0.

For details on the model, see the main model card.

License

Apache 2.0

Downloads last month
756
GGUF
Model size
2B params
Architecture
qwen35
Hardware compatibility
Log In to add your hardware

4-bit

5-bit

6-bit

8-bit

16-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for moganai/MoganOCR-TR-GGUF

Finetuned
Qwen/Qwen3.5-2B
Quantized
(6)
this model

Collection including moganai/MoganOCR-TR-GGUF