Skip to content
AI Model Radar

olmOCR 7B 0225 preview

allenai/olmOCR-7B-0225-preview
Open Source · chat · open-weights
GA Alert me on changes
Context
16K
Max output
—
Input $/1M
$0.27
Output $/1M
—
Modalities
text · image
Released
15 Jan 2025
License: apache-2.0 · allenai/olmOCR-7B-0225-preview
AI summary
● machine-written

Allen AI releases olmOCR 7B preview model for document OCR

olmOCR 7B 0225 preview is a 7-parameter vision language model fine-tuned from Qwen2-VL-7B-Instruct for optical character recognition in PDFs and documents. The model processes both text and image inputs and is distributed under the Apache 2.0 license with open weights via Hugging Face.

What's new
  • Fine-tuned from Qwen2-VL-7B-Instruct using olmOCR-mix-0225 dataset
  • Achieves 70.4% on MDPBench overall benchmark
  • Supports efficient inference via sglang toolkit for document-scale processing
  • Multiple quantization formats available (GGUF: Q4_K, Q6_K, Q8_0 and others)
Best for
Optical character recognition in born-digital and photographed PDFsDocument content extraction at scaleMulti-language document processing (Latin, German support evidenced)Local deployment via quantized formats
Sources

Source: https://huggingface.co/allenai/olmOCR-7B-0225-preview