LightOnOCR-3
#29 in Open-Source-SpraakmodelleLighton · v3 · siet 2026-10-08 · 2× · tolest 09. Okt. 2026
32
Momentum
LightOnOCR-3 is a family of open-weight document intelligence models from LightOn (Paris), available in 0.8B, 1B and 4B sizes under the Apache 2.0 license. The models transcribe text, output layout with bounding-box coordinates, describe images and extract data from charts. The models are intended for deployment on the user's own infrastructure.
Momentum-Verloop
11.07.09.10.
Features
| License | Apache 2.0 (research and commercial use) |
| Multimodality | Image-to-text (vision-language); outputs text, layout/bounding boxes, image descriptions and chart data |
| Platform | Open-weight models on Hugging Face; served via vLLM on own infrastructure |
| Release Date | October 8, 2026 |