cstr/lightonocr-crispembed-GGUF

LightOnOCR-2-1B — CrispEmbed GGUF

LightOnOCR-2-1B (LightOn, Apache-2.0): Pixtral ViT (24L, 1024d) + Qwen3 decoder (28L, 1024d). OCR Arena #2. 1B params.

Files

File	Type	Size	Notes
lightonocr-1b-f16.gguf	F16	2.3 GB	Full precision
lightonocr-1b-q8_0.gguf	Q8_0	1.1 GB	Best quality quantized
lightonocr-1b-q4_k.gguf	Q4_K	0.6 GB	Desktop/edge target

Usage

# CLI
crispembed --ocr lightonocr-1b-q8_0.gguf image.png

# Server
crispembed-server --ocr lightonocr-1b-q8_0.gguf
curl -X POST http://localhost:8080/math/ocr -d '{'"image": "document.png"}'

from crispembed import CrispMathOcr
ocr = CrispMathOcr("lightonocr-1b-q8_0.gguf")
text = ocr.recognize("document.png")
conf = ocr.mean_confidence
print(f"{text} (confidence: {conf:.2f})")

Source

• Original: LightOn/LightOnOCR-2-1B
• Converted with CrispEmbed convert-lightonocr-to-gguf.py

Provenance and EU AI Act Art. 53 note

• Upstream model: LightOn/LightOnOCR-2-1B.
• Upstream licence: apache-2.0. This repository redistributes under the same terms; it grants no rights the upstream licence does not.
• What was done here: format conversion and/or quantisation only (GGUF). No training, no fine-tuning, no merging, no distillation, no change to architecture, vocabulary or capability. Only the numeric representation of the upstream weights differs.
• Training data: documented — where it is documented at all — by the upstream provider; see the upstream model card. No training data was used, added or selected by this repository.
• Provider status: under Regulation (EU) 2024/1689 the upstream authors remain the provider of this model. Converting the serialisation format does not make this repository the provider of a new general-purpose AI model, and no such claim is made. Questions about training content, copyright policy or model capability belong upstream.