LibreDeepSeekOCR1

DeepSeek-OCR (3B mixture-of-experts document OCR model) weights mirrored for LibreYOLO's ocr task.

Source

Mirrored from deepseek-ai/DeepSeek-OCR at commit 9f30c71f441d010e5429c532364a86705536c53a (code, paper arXiv:2510.18234). Copyright (c) 2023 DeepSeek. Licensed under the MIT License.

Modifications

No learned parameters were changed. The safetensors, config.json, tokenizer and processor files are byte-identical to the upstream snapshot. The upstream remote-code Python files and example assets are not included: LibreYOLO runs its own PyTorch port of the architecture and never executes remote code.

Usage

from libreyolo import LibreVLM

model = LibreVLM("deepseek-ocr")              # grounded OCR, Gundam resolution
result = model.predict("page.png")
print(result.ocr.texts)                       # one transcript per text line

page = LibreVLM("deepseek-ocr", prompt="markdown").predict("page.png")
print(page.ocr.markdown)

License

MIT License. See LICENSE and NOTICE.