Qianfan OCR vs Chandra
Pick Qianfan OCR if you self-host and want a permissive Apache-2.0 licence with no revenue cap, plus 192 languages and handwriting in one 4B model. Pick Chandra if you want a hosted API with per-page pricing and the higher published olmOCR-bench score (85.8 vs 79.8), and your company is under $2M in funding or revenue or will buy a licence.
Qianfan OCRbaidu/qianfan-ocr | Chandradatalab/chandra | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 2.0 s / page | 2.0 s / page |
| Best for | Chinese documents | Complex layouts and handwriting |
| Inputs | PDF, images | PDF, images |
| Languages | Chinese, English and more | 40+ languages |
| Max pages per request | 200 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | ✓ Yes |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | No | ✓ Yes |
| Confidence | No | No |
| Forms | No | ✓ Yes |
| Multilingual | ✓ Yes | ✓ Yes |
| On OCRRouter | Available | Available |
Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.
Which to choose
Choose Qianfan OCR if
- Apache-2.0 allows commercial self-hosting at any company size, with no revenue threshold.
- It covers 192 languages, handwriting and scene text, and outputs Markdown, JSON, HTML, LaTeX and bounding boxes.
- It scores 93.12 on OmniDocBench v1.5, the top end-to-end score in Baidu's table, and suits key information extraction, tables and formulas.
Choose Chandra if
- The hosted Datalab API has clear per-page pricing: $1, $3 or $7.50 per 1,000 pages, with $20/month free usage.
- It scores 85.8 on olmOCR-bench and averages 77.8% across 43 languages, which suits multilingual scans and PDF to Markdown for RAG.
- It returns layout chunks with bounding boxes and labels per block, and the launcher has presets for A100, L40S and RTX 4090/5090.
Key differences
| Qianfan OCR | Chandra | |
|---|---|---|
| Licence | Apache-2.0, commercial use allowed | Code Apache-2.0; weights modified OpenRAIL-M, free under $2M funding/revenue, paid licence above that |
| How you run it | Self-host open weights (benchmarked on one A100 with vLLM) or Baidu Qianfan platform | Self-host via vLLM or Hugging Face, or hosted Datalab API |
| Price | Free self-hosted, you pay for compute; hosted price on Baidu Qianfan not published here | Self-hosted free within licence terms; API $1, $3 or $7.50 per 1,000 pages |
| Output formats | Markdown, JSON, HTML, LaTeX, bounding boxes | Markdown, HTML, JSON, bounding boxes |
| Benchmark | 93.12 OmniDocBench v1.5; 79.8 olmOCR-Bench (authors) | 85.8 olmOCR-bench (Chandra 2, authors); OmniDocBench not published |
| Speed | 1.024 pages/s on an A100 with W8A8; thinking mode can emit up to 16,384 tokens | About 1.44 pages/s on one H100 at 96 concurrent sequences; one reviewer measured 66 s per page on an RTX 4090 with the Hugging Face path |
| Known failure | Slower than MinerU 2.5; not suited to CPU-only deployment | Token-loop hallucinations reported on Russian/Kazakh scans; weak on Urdu (43.2%), Sanskrit (51.1%) and old scans |
Sources
- Qianfan-OCR model card
- Chandra GitHub README
- Datalab pricing
- GitHub issue #51 token loops
- Datalab anti-hallucination blog
What each one is
Qianfan OCR
Baidu's OCR-tuned vision model. Strong on Chinese documents and tables.
Qianfan OCR detailsChandra
Datalab's OCR model that keeps layout and handles forms, tables and handwriting.
Chandra detailsTry both on your own file
Run the file once with each model and compare the output.
PDF, PNG, JPG, WebP, HEIC. Up to 20 MB each, 10 files.No file?
Use one, fall back to the other
{"model": "baidu/qianfan-ocr","fallbacks": ["datalab/chandra"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Qianfan OCR vs SuryaQianfan OCR vs olmOCRPaddleOCR-VL vs ChandraJina OCR vs ChandraChandra vs RolmOCRDeepSeek-OCR vs Qianfan OCRQianfan OCR vs OCR.spaceChandra vs Pixtral
Qianfan OCR alternativesChandra alternativesBest OCR models in 2026