Tencent's WeVisDoc-4B hits 95.38 on OmniDocBench as two new OCR models land on Hugging Face

NielsRogge · x · 2026-09-18

Two new OCR models landed on Hugging Face: Tencent's WeVisDoc (2B and 4B, Apache 2.0) and jinaai/jina-ocr-v1 (non-commercial license). Per the OmniDocBench v1.6 benchmark on Papers with Code, WeVisDoc-4B leads with an overall score of 95.38, though the current SOTA remains NaviDC-OCR.

Related event: Tencent WeVisDoc and Jina's jina-ocr-v1 open-source OCR models launch same day(5 posts)→

Original post →

More from Models

Models channel →