TeleOCR needs about 2.2 GB of memory at its main file (a 1.5 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 2 quants, from 1.5 GB to 2.0 GB (Q8_0).
TeleOCR – lightweight vision‑language model for document parsing
TeleOCR is a ~1.2B parameter model designed for document parsing across digital and camera‑captured documents. It unifies layout, text, LaTeX formulas, tables, and code extraction in a single framework. The model is open‑source under Apache‑2.0 and supports llama.cpp via GGUF conversion.
Good at
- Document layout parsing
- Text, formula, table, and code extraction
- Robustness to distorted documents
Limits
- No vision support in llama.cpp
- Requires Qwen3‑style decoder, not compatible with all backends
- Peak VRAM ~6.4 GB at batch 8
Best for
- OCR pipelines for mixed media documents
- Batch document analysis on GPUs
- Research on document understanding
Who it’s for. Developers building OCR systems that need a single model for both digital and camera‑captured documents. Researchers requiring a lightweight, open‑source baseline for document parsing tasks.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 1.5 GB | 2.2 GB | Fits | Get file → | |
| Q8_0 | 2.0 GB | 2.7 GB | Fits | Get file → |
For 8‑GB VRAM, the Q8_0 GGUF file (1.97 GB) is recommended; it uses about 2.7 GB of VRAM at 8k context. The unknown quant file (1.48 GB) uses about 2.2 GB of VRAM at 8k context and may fit tighter budgets. Memory estimates are approximate.
What GPU do you need for TeleOCR?
These figures add the file size, an fp16 KV cache at an 8K context window 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 2.7 GB in use).
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about TeleOCR
How much VRAM does TeleOCR need?
About 2.2 GB at its main file 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 TeleOCR on an 8 GB GPU?
Yes, with Q8_0: the 2.0 GB file needs about 2.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of TeleOCR GGUF?
It depends on the quant: file is 1.5 GB, Q8_0 is 2.0 GB.
Where can I download TeleOCR GGUF?
The files are on Hugging Face in the rleo/TeleOCR-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 TeleOCR 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 TeleOCR on 8 GB VRAM?
Yes. The Q8_0 file uses about 2.7 GB of VRAM at 8k context, well below 8 GB. Peak VRAM is ~6.4 GB at batch 8 on an RTX 4090.
Is TeleOCR compatible with llama.cpp?
TeleOCR’s decoder is Qwen3‑style and not supported by llama.cpp. Use the teleocr‑rs GGUF layout instead.