Baidu Wenxin Releases PaddleOCR-VL-1.6: Accuracy Exceeds 96.33%, Refreshing Document Parsing SOTA
Baidu's Wenxin has released PaddleOCR-VL-1.6, a new version of its open-source OCR tool. This update achieves over 96.33% accuracy on the OmniDocBench v1.6 benchmark, surpassing major models like Gemini-3-Pro and GPT-5.2. The model demonstrates significant improvements in understanding complex documents, including scanned papers, bent documents, and screen captures, making it a leading solution for document digitization. AI
IMPACT Sets new SOTA on document parsing benchmarks, potentially accelerating enterprise adoption of advanced OCR solutions.