LlamaParse vs Surya
Pick LlamaParse if you want a hosted API that turns mixed PDF and office files into Markdown or JSON for RAG, and you can validate the output. Pick Surya if you want to run OCR yourself, need bounding boxes and reading order, or want to keep documents on your own hardware. The main split is hosted and per-page priced versus self-hosted and free below $5M funding or revenue.
LlamaParsellamaindex/llamaparse | Suryadatalab/surya | |
|---|---|---|
| Type | OCR API | Open source |
| Price | 5 credits / page | 2 credits / page |
| Per 1,000 pages | $5.00 | $2.00 |
| Typical speed | 3.0 s / page | 1.5 s / page |
| Best for | RAG ingestion | Layout and line-level OCR |
| Inputs | PDF, images, DOCX | PDF, images |
| Languages | 80+ languages | 90+ languages |
| Max pages per request | 1,000 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | ✓ Yes |
| Confidence | No | No |
| Forms | No | No |
| 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 LlamaParse if
- It takes PDFs and office files in one service, so you do not run any model yourself.
- Tiers run from Fast at 1 credit per page to Agentic Plus at 45, about $1.25 to $56 per 1,000 pages, so you can pay more only for hard layouts.
- It includes 10,000 free credits per month, and re-parsing the same file within 48 hours is free because results are cached.
Choose Surya if
- It is free to self-host (you pay for compute), with Apache-2.0 code and weights free for research, personal use and companies under $5M funding or revenue.
- One model returns layout labels, reading order, OCR, HTML tables and KaTeX math with per-block bounding boxes, for 90+ languages.
- It runs on CPU and Apple Silicon via llama.cpp with peak memory of about 4.3 to 5.9 GB on an M4 Max, or on an NVIDIA GPU via vLLM at 4.52 pages/s on one H100.
Key differences
| LlamaParse | Surya | |
|---|---|---|
| How you run it | Hosted API from LlamaIndex (LlamaCloud) | Self-host: a lightweight client against a vLLM or llama.cpp server you run |
| Licence | Proprietary | Code Apache-2.0; weights under modified AI Pubs Open Rail-M, commercial licence needed above $5M funding or revenue |
| Price | $1.25 per 1,000 credits; about $1.25 to $56 per 1,000 pages by tier; 10K free credits per month | Free, you pay for compute |
| Output formats | Markdown, JSON, plain text, images | JSON, HTML, LaTeX, bounding boxes, plain text |
| Benchmarks | Not published | 83.7% on olmOCR-bench (Surya 2.5), reported by the authors |
| Speed | Premium mode took 4 to 5 minutes in one reported job (GitHub #623) | 4.52 pages/s on one H100 with vLLM; 0.198 pages/s on an M4 Max with llama.cpp Q8_0 |
| Known failure | Invented transactions on bank statements (#450); hallucinates on low-quality scans with no confidence score (#483) | Multilingual tables score 58.1; Surya 2.5 table output lacks row and column geometry; not for handwriting or photos |
Sources
- LlamaIndex pricing
- GitHub issue #450
- GitHub issue #483
- GitHub issue #623
- Surya GitHub README
- Datalab pricing and license FAQ
What each one is
Surya
Open-source OCR toolkit with line detection, layout analysis, reading order and tables.
Surya 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": "llamaindex/llamaparse","fallbacks": ["datalab/surya"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
LlamaParse vs dots.ocrLlamaParse vs ChandraMathpix Convert vs SuryaClaude vs SuryaLlamaParse vs Reducto ParseLlamaParse vs DoclingGemini Flash vs SuryaSurya vs dots.ocr
LlamaParse alternativesSurya alternativesBest OCR models in 2026