# MonkeyOCRv2 Sets an Open Document-Parsing Record With an 11x Smaller Encoder

Trained on 113 million document images across 17 languages, the frozen encoder paired with a 0.7-billion-parameter language model tops MDPBench by 2.8 points absolute.

- Published: 2026-07-15T05:32:00.431Z
- Canonical: https://polylog.news/ai/2026-07-15/monkeyocrv2-sets-an-open-document-parsing-record-with-an-11x
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [Polylog editors](https://polylog.news), [arXiv](https://arxiv.org/abs/2607.11562)

Researchers from Huazhong University of Science and Technology and Kingsoft Office released MonkeyOCRv2, a visual-text foundation model for document AI built on MonkeyDoc v2, which they describe as the largest document-image pretraining cor…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-07-15/monkeyocrv2-sets-an-open-document-parsing-record-with-an-11x (subscription information: https://polylog.news/pricing).