Alibaba's 0.8B OvisOCR3 tops MDPBench at 88.8, edging out Gemini-3-pro
NielsRogge · x · 2026-10-30
Alibaba released OvisOCR3, an OCR model that scores 88.8 on the MDPBench document-parsing benchmark, ahead of Gemini-3-pro-preview at 88.4. Weights are available on Hugging Face.
Despite its compact 0.8B size, the model unifies five task families: document parsing, general OCR, key information extraction, figure captioning, and cross-page table reconstruction. Training combines domain-balanced mid-training, multi-teacher on-policy distillation from five 4B domain teachers, and mixed-task RL with domain-specific rewards. It hits 96.74 on OmniDocBench v1.6, ranks first on OCRBench v2 (65.9 English / 61.3 Chinese), and leads on PureDocBench, CrossTabBench, and more.
Related event: Alibaba releases 0.8B OvisOCR3, tops MDPBench leaderboard(2 posts)→
More from Models
- Users complain Opus 5.5 feels heavily nerfed — RileyRalmuto · 2026-10-31
- Bug bounty researcher flags AI security platform sessions as potential training data of unpatched exploits — npinto · 2026-10-31
- MOSS AMA Roadmap: Full-Duplex Live Model, Image Turbo, Consumer-GPU Text2Vid — tomByrer · 2026-10-31
- Grok says rumored Gemini 4 Carbon cyber model is wrapped in guardrails, not leaked carelessly — AccBalanced · 2026-10-31
- Alleged 8TB "Claude Fable 5" weight leak from AWS S3 looks fabricated, teardown finds — AccBalanced · 2026-10-31
- NVIDIA ships NVFP4 DeepSeek-V4.1-Flash, 81GB smaller with near-identical quality — AccBalanced · 2026-10-31