5 million parameters level the scale of billion-level large models: Baidu PaddleOCR surpasses Tesseract to top the GitHub OCR charts

BlockBeatNews · 2026-03-30T08:52:04+00:00

Baidu PaddlePaddle's open-source OCR toolkit PaddleOCR has surpassed Tesseract with 73,300 GitHub stars, making it the highest-starred OCR project. PaddleOCR supports over 100 languages, and recent updates have enhanced its performance and accuracy.

BlockBeatNews

2026-03-30 08:52:04

Abstract generation in progress

According to monitoring by 1M AI News, Baidu’s open-source OCR toolkit PaddleOCR has surpassed Google’s long-established OCR engine Tesseract (73,200 stars) with 73,300 stars on GitHub, making it the highest-rated OCR project on GitHub. The third-ranked MinerU has 57,500 stars. PaddleOCR was open-sourced in 2020, supporting over 100 languages and covering more than 160 countries and regions.

PaddleOCR has recently undergone intensive updates, with the release of PP-OCRv5 last week featuring only 5 million parameters, achieving accuracy comparable to that of billion-parameter visual language models on standard OCR benchmarks; PaddleOCR-VL-1.5 set a new record with an accuracy of 94.5% on the document parsing benchmark OmniDocBench v1.5.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.