Xiaomi OCR 0 Extraction Desk
Point the 0.8B engine at any page and pull out text, tables and math
None

None

None

Long Story Video Skill

Turn your idea into 10s–10min video—AI writes scripts, prompts, and generates footage automatically.

Ads Video Skill

Ads Video Skill

Generate professional ads and sales videos—AI auto-generates scripts, prompts, and footage.

3D Science Explainer Video Skill

Convert scientific concepts into stunning 3D explain animations

AI Video Prompt Generator

Feedback

freeTrialImage.bannerPity

freeTrialImage.upgradeUnlock

  • ✓freeTrialImage.benefitHd
  • ✓freeTrialImage.benefitWatermark
  • ✓freeTrialImage.benefitUnlimited

xiaomi ocr 0

Scan pages, tables, formulas or handwriting and get tidy structured text from the 0.8B Xiaomi OCR 0 engine — no signup, results in seconds.

All Tools

Discover our comprehensive AI-powered animation toolkit

Xiaomi OCR 0: Small Model, Serious Document Reading

Parsing, visual question answering and key-field extraction all run on the same compact checkpoint, so you can retire several single-purpose tools at once.

  • One Checkpoint, Three Jobs
    Reading pages, answering questions about them and pulling fields all happen inside a single lightweight model.
  • Benchmark Leader
    It ranks first on Real5-OmniDocBench with 95.24 and takes second on OmniDocBench v1.6 and Wild-OmniDocBench.
  • Holds Up on Messy Scans
    Warped, skewed and phone-captured pages are handled as reliably as flat digital files.

Run Xiaomi OCR 0 in Three Quick Steps

From a raw page image to tidy structured text in well under a minute.

Core Strengths of Xiaomi OCR 0

Light on resources yet strong on messy scans, handwriting and open-weight deployment — here is what this engine actually delivers.

Full-Page Parsing

Works on ordinary pages as well as scanned, warped, tilted and re-photographed ones.

Table & Equation Reading

Scores 97.84 on UniMER-Test for formulas and 92.27 on PubTabNet for table structure.

Handwriting Support

Reaches 73 on Chronicles-OCR for mature scripts, matching Seed 2.0 Pro.

Question Answering on Pages

Averages 83.2 across five VQA suites, beating a 2B competitor at a fraction of the parameter count.

Structured Field Capture

Reads receipts and forms into organised fields rather than flat transcription.

Runs Locally, Openly

Apache-2.0 weights ship with GGUF and BF16 builds for llama.cpp, Ollama and Transformers.

FAQ

Xiaomi OCR 0: Questions Answered

Quick answers on what this 0.8B model does, how it scores and how to run it offline.

1

What exactly is this model?

It is a 0.8B vision-language checkpoint tuned for OCR: it reads a document and answers questions about it in a single pass.

2

Where does it rank on benchmarks?

First on Real5-OmniDocBench, second on OmniDocBench v1.6, and second again on Wild-OmniDocBench.

3

Does handwriting work?

Yes — clerical, cursive, running and regular calligraphy are all covered, scoring 73 on Chronicles-OCR.

4

What does the training recipe look like?

Roughly 170M OCR-heavy samples, combined with Q-Mask text anchoring, continued pretraining and mixed-task reinforcement learning.

5

Can it run offline?

Yes, Apache-2.0 weights are public, and community GGUF and BF16 conversions target llama.cpp, Ollama and Transformers.

6

Is transcription the only output?

No — it also answers questions about a page and returns key fields as structured data.

Put Xiaomi OCR 0 to Work on Your Files

Upload a single page image and watch parsed text, tables and structured fields come back in moments — free to start.