Chemistry · math · tables
01 / SCIENCE
Scientific layout understanding
Purpose-built for chemical formulas, mathematical notation, complex tables, and multi-column layouts — preserving reading order and logical structure.
MinerU Online · No install, open and go
Make every document
machine-readable
Turn PDFs, papers, and web pages into clean Markdown, HTML, and structured JSON. Tuned for scientific layouts, chemical formulas, and complex tables — built for AI agents and RAG pipelines.
- Pages parsed daily
- 1.2M+
- Pages parsed daily
- Supported formats
- 30+
- File & web formats
- Service uptime
- 99.9%
- Service uptime
paper_3a.pdf → output.md
Scheme 2. Synthesis of compound 3a
J. Med. Chem. 2026
Figure 2. Structure of 2,4,6-trinitrophenol
## 2.2 Synthetic route
Target product **3a** parsed as:
- SMILES `O=[N+]([O-])C1=CC([N+](=O)[O-])=C(O)C([N+](=O)[O-])=C1`
- Formula $C_6H_3N_3O_7$
- Confidence 0.98 · page 7
Structure accuracy 96.7%
1.8s elapsed
Drop a file here, or click to browse
Batch upload · up to 50 MB per file · original layout preserved
Output
Tasks
3 tasks| Task | Status | Type | Model | Created | Actions |
|---|---|---|---|---|---|
|
NeurIPS-2024-paper.pdf
|
Completed | Document | MinerU XL | 2026/09/17 14:24 | |
|
arxiv.org/abs/2409.xxxxx
|
Completed | Web | MinerU HTML | 2026/09/17 13:50 | |
|
organic-chemistry-lab-manual.pdf
|
Parsing | Document | MinerU XL | 2026/09/17 14:58 | |
|
polymer-characterization-report.docx
|
Completed | Document | MinerU XL | 2026/09/17 12:07 | |
|
nature.com/articles/s41586-024-07…
|
Completed | Web | MinerU HTML | 2026/09/17 09:32 |
No tasks yet
Upload a file or paste a web link — results will show up here.
Capabilities
A parsing engine built for complex documents
Beyond text extraction. MinerU understands the structure, semantics, and hierarchy of scientific layouts.
Semantics · bbox · hierarchy
02 / AGENTS
Native to AI agents
Structured JSON with semantic tags for headings, paragraphs, tables, figures, and formulas — plug directly into RAG, knowledge bases, and agent workflows.
100 pages · 1.8s total
03 / SPEED
High-speed elastic compute
Elastic cloud scheduling parses hundred-page documents in seconds, while batch jobs run in parallel to meet enterprise throughput.
Who uses MinerU
Built for research, engineering, and business teams
From one person's reading list to an organisation-wide knowledge base, MinerU sits at the first gate where documents enter your systems.
Researchers
Scientific researchers
Turn papers, patents, and lab notes into searchable structured text in bulk. Formulas and chemical notation survive intact, so literature reviews stop being manual transcription.
Developers
AI application developers
One line of code. Get semantically tagged JSON straight into your RAG pipeline, knowledge base, or agent workflow — no parsing pipeline of your own to maintain.
Enterprise
Enterprise & institutional teams
Batch-ingest research reports, contracts, and archives. Pair it with on-premise deployment and usage reporting to make document assets genuinely searchable.
Parsing in action
See what comes back from a real page
Handed the same document, a generic OCR gives you a wall of text. MinerU keeps the structure, the chemistry, and the math.
Molecular diagram → searchable, computable chemistry
Structures are no longer thrown away as images. MinerU reads ring systems and substituents, then emits SMILES, molecular formulas, and LaTeX.
# structure → structured chemistry
{
"type": "chemical_structure",
"smiles": "O=[N+]([O-])C1=CC([N+](=O)[O-])=C(O)C([N+](=O)[O-])=C1",
"formula": "C6H3N3O7",
"iupac": "2,4,6-trinitrophenol",
"latex": "\ce{(O2N)3C6H2OH}"
}
# inline Markdown output
2,4,6-trinitrophenol (picric acid, $C_6H_3N_3O_7$, SMILES: O=[N+]([O-])C1=CC([N+](=O)[O-])=C(O)C([N+](=O)[O-])=C1)
Generic OCR
The diagram is dropped as an image — all that survives is "see Fig. 2". No search, no computation.
MinerU
Recovered as SMILES / formula / LaTeX — ready to index, query, and run property calculations on.
Formulas → clean LaTeX
Subscripts, radicals, fractions and Greek letters survive intact — output compiles as-is.
∫0∞ e−x² dx = √π2
π²6 = ∑n=1N 1n²
# standard LaTeX, renders as-is
\int_{0}^{\infty} e^{-x^{2}}\,dx
= \frac{\sqrt{\pi}}{2}
\frac{\pi^{2}}{6}
= \sum_{n=1}^{N} \frac{1}{n^{2}}
# human-readable text kept too
∫₀^∞ e^(−x²) dx = √π / 2Merged headers & spanning tables → structured rows
Multi-level headers and tables that continue across pages are stitched back together, with cell coordinates intact.
# merged headers flattened to Markdown
| Group | Control mean | Control SD | Treatment mean | Treatment SD |
| --- | ---:| ---:| ---:| ---:|
| Sample A | 12.4 | 0.83 | 18.9 | 1.02 |
| Sample B | 14.1 | 0.91 | 21.3 | 1.14 |
| Sample C | 15.7 | 1.05 | 24.6 | 1.28 |
| Sample D | 17.2 | 1.11 | 26.8 | 1.35 |
# coordinates & merges kept in JSON
{"rows": 5, "cols": 5, "merged": [2, 2]}API reference
Integrate in one line, call it like a local function
A RESTful API and multi-language SDKs cover upload, status, and download — with clear error codes and webhooks so integration stays painless.
- RESTful endpoints with cURL / Python / Node.js / Go examples
- Choose synchronous responses or async jobs
- Structured error codes with retry guidance
- Webhook callbacks on completion, no polling needed
Python
Node.js
cURL
# pip install mineru
import mineru
client = mineru.Client(api_key="your-api-key")
job = client.parse.upload(
file=open("paper.pdf", "rb"),
output_format="markdown",
model="mineru-xl"
)
# Wait for the job and save the result
result = job.wait()
result.save("output.md")
print(result.pages, "pages parsed")Price
Pick by usage, switch anytime
From individual researchers to enterprise teams. Every plan includes online parsing and result export.
Free
Sign up and start right away
$0/ month
No payment method required
- 50 free pages every month
- PDF, images, and web links
- Markdown / HTML export
- 3 concurrent tasks
- Community support
Pro · Monthly
Billed monthly, cancel anytime
$9.9/ month
Monthly billing, cancel anytime
- 10,000 pages every month
- Priority access to MinerU XL
- Structured JSON and full API access
- Batch jobs, concurrency, and webhooks
- Email support
Pro · Annual
One payment, best value for long-term use
$59.9/ year
About $4.99 / month — save 50%
- 10,000 pages every month
- Everything in the monthly plan
- Annual usage reports and invoicing
- Dedicated success manager, priority response
- Early access to new models and features
On-premise
Your data never leaves your premises
For finance, healthcare, government, and research institutions, MinerU deploys fully on-premise: models and data run entirely inside your own network.
100%
Data stays local
Compliant
Security framework ready
24/7
Dedicated support
Customer stories
- A top-tier securities research institute Private parsing for an internal research knowledge base — 50,000 pages a day with zero data egress.
- A tertiary hospital research department Structured extraction from medical literature and trial reports, fully compliant with research data rules.
- A provincial government cloud Digitized archives of official documents powering intelligent search and Q&A.
