Chandra vs DeepSeek-VL2
Pick Chandra to turn scanned pages into Markdown, HTML or JSON with bounding boxes, especially multilingual documents or RAG ingestion. Pick DeepSeek-VL2 to answer questions about a single document or chart page, where its DocVQA score of 93.3 fits. The main split is the job: Chandra transcribes whole pages, DeepSeek-VL2 understands one page at a time.
Chandradatalab/chandra | DeepSeek-VL2deepseek/deepseek-vl2 | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 2.0 s / page | 3.0 s / page |
| Best for | Complex layouts and handwriting | Charts and tables |
| Inputs | PDF, images | images |
| Languages | 40+ languages | Mainly English and Chinese |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | No |
| Forms | ✓ Yes | No |
| Multilingual | ✓ Yes | No |
| 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 Chandra if
- It outputs Markdown, HTML and JSON plus layout chunks with bounding boxes and labels per block, which suits PDF-to-Markdown for RAG.
- It scores 85.8 on olmOCR-bench and averages 77.8% across 43 languages, so it fits multilingual scans.
- You can self-host it or call the hosted Datalab API at $1, $3 or $7.50 per 1,000 pages, with a free $20/month tier.
Choose DeepSeek-VL2 if
- It scores 93.3 on DocVQA and 86.0 on ChartQA, which fits question answering on documents and charts.
- The weights allow commercial use with no revenue or funding cap.
- The Tiny model runs under 40GB of GPU memory, and only 1.0B to 4.5B parameters are active per token.
Key differences
| Chandra | DeepSeek-VL2 | |
|---|---|---|
| Licence | Code Apache-2.0; weights modified OpenRAIL-M, free for research, personal use and companies under $2M funding or revenue | Code MIT; weights under the DeepSeek Model License, commercial use allowed |
| How you run it | Self-host with vLLM or Hugging Face, or use the hosted Datalab API | Self-host only |
| Price | Self-hosted is free within licence terms. Hosted API is $1, $3 or $7.50 per 1,000 pages | Free, self-hosted; you pay for compute |
| Output formats | Markdown, HTML, JSON, bounding boxes | Plain text, Markdown |
| Benchmark | olmOCR-bench overall 85.8 (Chandra 2) | DocVQA 93.3; OCRBench 834 (Small) |
| Hardware | Single NVIDIA GPU via vLLM; benchmarks on one H100 80GB, with presets for A100, L40S and RTX 4090/5090 | Small may need 80GB, or 40GB with incremental prefilling; Tiny runs under 40GB |
| Known failure | Token loops on some Russian and Kazakh scans; weak on Urdu (43.2%) and old scans | Struggles with blurry images; supports only a few images per session |
Sources
What each one is
Chandra
Datalab's OCR model that keeps layout and handles forms, tables and handwriting.
Chandra detailsDeepSeek-VL2
DeepSeek's mixture-of-experts vision models (December 2024) with OCR and table and chart understanding.
DeepSeek-VL2 detailsTry both on your own file
Run the file once with each model and compare the output.
Use one, fall back to the other
{"model": "datalab/chandra","fallbacks": ["deepseek/deepseek-vl2"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Chandra alternativesDeepSeek-VL2 alternativesBest OCR models in 2026