OvisOCR2
OvisOCR2 is a compact 0.8B parameter end-to-end model for page-level document parsing. Given a document page image, it generates a Markdown representation in natural reading order, covering text, formulas, tables, and visual regions. Built by post-training Qwen3.5-0.8B with a multi-stage SFT, RL, and OPD training recipe.
- Type
- Multimodal
- Weights
- Open
- Downloads
- 198k
- Likes
- 473
- Pipeline
- image text to text
- License
- apache-2.0
Model hub
ATH-MaaS/OvisOCR2
- Repo
- ATH-MaaS/OvisOCR2
- Library
- transformers
- Access
- Public
Technical tags and languages
Papers
1
More from Alibaba
14
-
727 DL
ABot World 0 5B LF
Video
ABot-World-0-5B-LF is a 5B parameter causal video world model finetuned from Wan2.2-TI2V-5B. It generates inf…
-
FunAudio ASR
Multimodal
FunAudio-ASR, also called Fun-ASR, is Tongyi Lab’s end-to-end speech recognition model family for production…
-
HappyHorse 1.0
Multimodal
HappyHorse-1.0 is a closed AI video generation model available through fal’s API and playground. It supports…
-
HappyHorse 1.1 T2V
Video
HappyHorse-1.1-T2V is a text-to-video generation model with improved semantic understanding, cinematic shot c…
-
HappyOyster 1.0
Video
Real-time interactive world model that generates immersive video environments from text and image prompts. Su…
-
Hunyuan HY-T1
Image
Hunyuan T1 is Tencent’s deep reasoning model positioned for stronger structured reasoning and long-context an…
-
78 DL
LOGOS-8B
Text
LOGOS-8B (Language Of Generative Objects in Science) is an 8B parameter autoregressive scientific foundation…
-
591 DL
Marco-o1
Text
Marco-o1 is Alibaba International Digital Commerce’s open reasoning model line aimed at real-world problem so…
About Alibaba
1
Chinese e-commerce and cloud leader behind Taobao, Tmall, and Alipay.