Woofun AI reports that Alibaba Cloud has released OvisOCR2, an open-source document parsing model built on Qwen3.5-0.8B. The system converts document pages directly into Markdown format, simultaneously restoring text, formulas, tables, and reading order without requiring separate layout analysis or content recognition steps.

OvisOCR2 achieved a comprehensive score of 96.58 on OmniDocBench v1.6, exceeding the pipeline model PaddleOCR-VL-1.6 which scored 96.33. This marks the first instance of an end-to-end model topping the leaderboard. The model is distributed under the Apache 2.0 license, with weights accessible via Hugging Face and Model Zoo.