Instructions to use noctrex/LightOnOCR-2-1B-ocr-soup-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use noctrex/LightOnOCR-2-1B-ocr-soup-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="noctrex/LightOnOCR-2-1B-ocr-soup-GGUF",
	filename="LightOnOCR-2-1B-ocr-soup-BF16.gguf",
)

llm.create_chat_completion(
	messages = "\"cats.jpg\""
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use noctrex/LightOnOCR-2-1B-ocr-soup-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M

Use Docker

docker model run hf.co/noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use noctrex/LightOnOCR-2-1B-ocr-soup-GGUF with Ollama:
```
ollama run hf.co/noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M
```

Unsloth Studio new

How to use noctrex/LightOnOCR-2-1B-ocr-soup-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for noctrex/LightOnOCR-2-1B-ocr-soup-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for noctrex/LightOnOCR-2-1B-ocr-soup-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for noctrex/LightOnOCR-2-1B-ocr-soup-GGUF to start chatting

Docker Model Runner
How to use noctrex/LightOnOCR-2-1B-ocr-soup-GGUF with Docker Model Runner:
```
docker model run hf.co/noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M
```

Lemonade

How to use noctrex/LightOnOCR-2-1B-ocr-soup-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull noctrex/LightOnOCR-2-1B-ocr-soup-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.LightOnOCR-2-1B-ocr-soup-GGUF-Q4_K_M

List all available models

lemonade list

Phenomenal model

by spanspek - opened Jan 24

Discussion

spanspek

Jan 24

This model is phenomenal.

Q8_0 and F32 mmproj, full context in under 4.5GB of VRAM and it runs incredibly quickly. I'm running this on a laptop with a RTX 4060 (8GB) and it flies

spanspek changed discussion title from Phenomenal to Phenomenal model Jan 24

nmkd

Feb 4

This is somehow more reliable than Qwen3-VL 8B or next-ocr, or pretty much any other model I've tried, and it's tiny at 1B.
Truly amazing model, managed to OCR an entire movie's SDH PGS subtitles in less than 2 mins with zero errors.

noctrex

Owner Feb 4

Yeah, the first one was good, but this second one is even better.

tertat

Mar 2

•

edited Mar 2

Used Q6_k . Just 2.13G and did unexpectedly well, used in lmstudio. Man where is the loss? How this performs too good while others need minimum 5gb for quality. Ran on ryzen 8600g igpu.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment