LightOnOCR-2-1B: a lightweight high-performance end-to-end OCR model family
•
76
AI & ML interests
None defined yet.
Recent Activity
Papers
View all Papers
Articles
Visual long document VLMs based on Mistral-Small-3.1-24B-Instruct-2503 and Qwen3-VL-32B-Instruct
The Case for End-to-End and Efficient Domain-Specific Vision-Language Models for OCR
-
lightonai/Reason-ModernColBERT
Sentence Similarity • 0.1B • Updated • 11k • • 206 -
lightonai/GTE-ModernColBERT-v1
Sentence Similarity • 0.1B • Updated • 14.3k • • 151 -
lightonai/answerai-colbert-small-v1
Sentence Similarity • 33.4M • Updated • 161 • • 3 -
lightonai/colbertv2.0
Sentence Similarity • 0.1B • Updated • 4.36k • • 4
French language models. These model were trained in early 2021 following the then scaling laws and using the exact same training data as the CamemBERT
LightOnOCR-2-1B: a lightweight high-performance end-to-end OCR model family
-
lightonai/LightOnOCR-2-1B
Image-Text-to-Text • 1B • Updated • 188k • 503 -
lightonai/LightOnOCR-2-1B-bbox
Image-Text-to-Text • 1B • Updated • 5.41k • 20 -
LightOnOCR 2 1B Demo
🐨93Extract and recognize text from images and PDFs
-
lightonai/LightOnOCR-2-1B-base
Image-Text-to-Text • 1B • Updated • 5.15k • 9
This collection gather datasets for embeddings pre-training and fine-tuning.
Bringing BERT into modernity via both architecture changes and scaling
-
answerdotai/ModernBERT-base
Fill-Mask • 0.1B • Updated • 860k • 992 -
lightonai/GTE-ModernColBERT-v1
Sentence Similarity • 0.1B • Updated • 14.3k • • 151 -
lightonai/Reason-ModernColBERT
Sentence Similarity • 0.1B • Updated • 11k • • 206 -
lightonai/modernbert-embed-large
Sentence Similarity • 0.4B • Updated • 25.5k • • 28
A suite of autoregressive generative models for protein sequences, with up to 1.2Bparameters, trained on over 280 million protein sequences.
900M models trained on 25BT to compare different data processing choices (filtering, sentence dedup, minhash, etc)
Visual long document VLMs based on Mistral-Small-3.1-24B-Instruct-2503 and Qwen3-VL-32B-Instruct
LightOnOCR-2-1B: a lightweight high-performance end-to-end OCR model family
-
lightonai/LightOnOCR-2-1B
Image-Text-to-Text • 1B • Updated • 188k • 503 -
lightonai/LightOnOCR-2-1B-bbox
Image-Text-to-Text • 1B • Updated • 5.41k • 20 -
LightOnOCR 2 1B Demo
🐨93Extract and recognize text from images and PDFs
-
lightonai/LightOnOCR-2-1B-base
Image-Text-to-Text • 1B • Updated • 5.15k • 9
The Case for End-to-End and Efficient Domain-Specific Vision-Language Models for OCR
This collection gather datasets for embeddings pre-training and fine-tuning.
-
lightonai/Reason-ModernColBERT
Sentence Similarity • 0.1B • Updated • 11k • • 206 -
lightonai/GTE-ModernColBERT-v1
Sentence Similarity • 0.1B • Updated • 14.3k • • 151 -
lightonai/answerai-colbert-small-v1
Sentence Similarity • 33.4M • Updated • 161 • • 3 -
lightonai/colbertv2.0
Sentence Similarity • 0.1B • Updated • 4.36k • • 4
Bringing BERT into modernity via both architecture changes and scaling
-
answerdotai/ModernBERT-base
Fill-Mask • 0.1B • Updated • 860k • 992 -
lightonai/GTE-ModernColBERT-v1
Sentence Similarity • 0.1B • Updated • 14.3k • • 151 -
lightonai/Reason-ModernColBERT
Sentence Similarity • 0.1B • Updated • 11k • • 206 -
lightonai/modernbert-embed-large
Sentence Similarity • 0.4B • Updated • 25.5k • • 28
French language models. These model were trained in early 2021 following the then scaling laws and using the exact same training data as the CamemBERT
A suite of autoregressive generative models for protein sequences, with up to 1.2Bparameters, trained on over 280 million protein sequences.
900M models trained on 25BT to compare different data processing choices (filtering, sentence dedup, minhash, etc)