OvisOCR2: A 0.8B Local Document Parsing Model
Sad_External6106 · reddit · 2026-07-13
A Reddit user introduced OvisOCR2, a locally runnable 0.8B end-to-end document parsing model based on Qwen3.5-0.8B.
It aims to convert full-page documents directly into structured Markdown, covering:
- Plain text
- Tables
- Formulas
- Reading order
The post highlighted benchmark results from the model card:
- OmniDocBench v1.6: 96.58
- PureDocBench: 75.06
Additionally, the weights are open-sourced under Apache 2.0 and support vLLM. The author noted its impressive performance for such a small footprint but emphasized the need for more real-world testing, asking the community if anyone has compared it against GLM-OCR, PaddleOCR-VL, MinerU, Chandra.
Related event: Alibaba Unveils OvisOCR2 for Local Document Parsing(4 posts)→
More from Multimodal
- Fake BTS of a dam disaster made with Dreamina Seedance for 29 cents — Standard_Spell5873 · 2026-10-27
- Redditor struggles with outfit swap: Qwen Image 2.1 breaks body proportions — Existing-Box-7866 · 2026-10-27
- Consistent characters in ComfyUI: RealVisXL V5 + img2img face reference workflow — MiaLarsonJL · 2026-10-27
- 29 cents gets you a Michael Bay rooftop shot with Seedance — PersonalSpace_4225 · 2026-10-27
- Moss Image Turbo 3-Step LoRA Tested: Fast Generation at 6-8 Steps — JadedScorpion · 2026-10-27
- Magnific teases October 28 release: 'We're making it move' — cuenca · 2026-10-27