Unstract Starter costs $499 a month for 5,000 pages and Growth costs $2,249 a month for 25,000 pages, with overage at $0.10 and $0.09 a page. After the 14-day trial you connect your own LLM, so the model provider bills you on top. For invoices that works out to roughly $76 to $110 per 1,000 pages all in, against $14.90 on a flat invoice plan with the model included.
PDF, JPG, PNG, BMP, HEIC and TIFF
Drop files here or click to upload
Up to 50 files
Uploading...
Unstract publishes two self-serve plans and a quote-only Enterprise tier. The page fee is only one of three costs you carry once the trial ends.
The trial includes Azure OpenAI GPT-4o with 1M tokens. After it, Unstract says you connect your preferred stack, which means your OpenAI, Anthropic, Bedrock or Azure account bills every extraction separately.
The smallest plan is $499 a month for 5,000 pages, or $416 a month paid yearly. A team processing 800 invoices a month pays for six times the volume it uses.
SOC 2 Type II, ISO 27001, HIPAA, a GDPR DPA, SSO and audit logs are marked as Enterprise only. Starter and Growth get encryption and standard support.
The toggle says save 25 percent. The listed prices, $416 against $499 and $1,874 against $2,249, work out to about 17 percent off on both plans.
Unstract is a platform. You define the schema and write the extraction prompts in Prompt Studio, test them, and deploy them as an API or ETL pipeline before the first invoice flows.
Pages that fail on the Unstract side are free. Pages that process are charged even when the extracted values are wrong, so a weak prompt costs the same as a good one.
Unstract sells two products on one pricing page. The Unstract platform is the extraction workspace with Prompt Studio, pipelines and review. LLMWhisperer is the OCR and text layout API underneath it, which you can also buy on its own. Prices are in US dollars.
Includes LLMWhisperer, Azure OpenAI GPT-4o with 1M tokens, an embedding model and a Postgres vector database, so you can extract on day one without wiring anything.
5,000 pages a month, overage at $0.10 a page, Agentic Prompt Studio, standard compliance and support. $416 a month billed yearly for 60,000 pages a year.
25,000 pages a month, overage at $0.09 a page. $1,874 a month billed yearly for 300,000 pages a year. Same feature set as Starter.
Custom volume and overage, SOC 2, ISO 27001, HIPAA, GDPR with a DPA, SSO, a 99.9 percent uptime SLA, cloud or self-hosted in your own VPC with unlimited pages.
Standalone parsing at $1, $5, $10 or $15 per 1,000 pages depending on mode, with 100 free pages a day. Included at no extra charge inside Unstract plans.
The core platform is on GitHub under the AGPL-3.0 license. Self-hosting removes the page fee but not the LLM bill, the servers or the engineering time.
Three numbers decide what Unstract really costs you for invoices.
Unstract bills pages, not documents. A two-page invoice is two pages. Use month end or quarter end volume, since every page over the plan bills at $0.09 or $0.10.
Pick the model you will connect and multiply its token rates by your prompts. Per-field prompting sends the document text once per field unless you turn on SinglePass extraction.
Someone writes and tests the schema, deploys the API or pipeline and maintains the prompts as vendors change layouts. Compare that total with a tool that already reads invoices.
Unstract is a capable, developer-friendly extraction platform. It is priced for teams that will build on it across many document types.
If you need documents flowing into a warehouse or vector database across contracts, statements and forms, Unstract ETL pipelines and connectors do work an invoice tool does not.
On-prem Enterprise runs inside your VPC with unlimited pages, and the AGPL edition can be audited line by line. That matters for healthcare and finance data rules.
If the job is vendor invoices into Excel, CSV or QuickBooks, paying $499 a month plus an LLM bill and owning prompts is the expensive route.
Under 5,000 pages a month you pay for capacity you never use. A flat invoice plan with the model included costs a fraction of it.
Unstract, built by Zipstack, is an LLM-powered platform for turning unstructured documents into structured data. You define the fields you want in Prompt Studio, test the prompts against sample documents, then deploy the result as an API or an ETL pipeline. Pricing is per page processed, with two self-serve plans and a quote-only Enterprise tier. These are the rates on the Unstract pricing page, in US dollars.
| Plan | Monthly billing | Yearly billing | Pages included | Overage | Compliance |
|---|---|---|---|---|---|
| Trial | $0 for 14 days | n/a | GPT-4o with 1M tokens bundled | n/a | Standard |
| Starter | $499 a month | $416 a month ($4,992 a year) | 5,000 a month or 60,000 a year | $0.10 a page | Standard |
| Growth | $2,249 a month | $1,874 a month ($22,488 a year) | 25,000 a month or 300,000 a year | $0.09 a page | Standard |
| Enterprise | Quote only | Quote only | Custom, unlimited on-prem | Custom | SOC 2, ISO 27001, HIPAA, GDPR |
Two details are easy to miss. First, the yearly plans meter pages per year rather than per month, so a quarter-end spike can borrow from quieter months instead of tripping overage. Second, the "save 25%" label on the billing toggle does not match the listed prices. $416 against $499 and $1,874 against $2,249 are both about 17 percent off, which is still worth having, just not a quarter.
On the page fee alone, Unstract costs $99.80 per 1,000 pages on Starter billed monthly and $83.20 on Starter billed yearly. Growth costs $89.96 per 1,000 pages billed monthly and $74.96 billed yearly. Overage is $100 and $90 per 1,000 pages. None of these figures includes the LLM, which you pay your model provider for separately.
The step from Starter to Growth pays off at a specific volume. Starter plus overage reaches the Growth price at 22,500 pages a month on monthly billing ($499 plus 17,500 pages at $0.10 equals $2,249). On yearly billing the crossover arrives a little earlier, at about 19,600 pages a month. Below that, stay on Starter and absorb the overage.
Page fee per 1,000 pages, US dollars, before any LLM bill
InvoiceExtractor figure is the Plus plan, $149 a month for 10,000 Base AI pages, with the AI model included in the price.
Only during the trial. The 14-day trial bundles Azure OpenAI GPT-4o with 1M tokens, an embedding model and a Postgres vector database. After that, the Unstract pricing page says you connect your preferred stack, so the LLM runs on your own provider account and every extraction adds a token charge on top of the page fee.
How big that charge is depends on the model and on how you prompt. As a rough guide, assume a one-page invoice turns into about 1,500 input tokens of LLMWhisperer text and about 500 output tokens of JSON. At published API rates that is roughly:
| Model you connect | Rate per 1M tokens (in / out) | LLM cost per 1,000 pages, one pass | All in on Starter, monthly |
|---|---|---|---|
| GPT-5 mini | $0.25 / $2.00 | about $1.40 | about $101 |
| GPT-5 | $1.25 / $10.00 | about $6.90 | about $107 |
| GPT-4o | $2.50 / $10.00 | about $8.75 | about $109 |
| Claude Sonnet 5 | $2.00 / $10.00 | about $8.00 | about $108 |
Those numbers assume one pass over the document. Prompt Studio lets you write one prompt per field, and in that mode the document text is sent with each prompt, so a 12-field invoice schema can multiply input tokens by up to 12. Unstract offers SinglePass extraction to send the document once for all fields, and it is the setting to check first if your token bill looks high. The page fee dominates either way: on Starter the LLM is usually under 10 percent of the total.
LLMWhisperer is the OCR and layout engine that prepares text for the LLM. Inside an Unstract plan it is included at no extra charge. Bought on its own it is pay as you go, with 100 free pages a day on every mode and no credit card needed to start.
| LLMWhisperer mode | Per 1,000 pages | Best for | Languages |
|---|---|---|---|
| Native Text | $1 | Digital PDFs with a text layer | All Unicode |
| Low Cost | $5 | Clean, high-quality scans | 120+ |
| High Quality | $10 | Medium or poor scans, phone photos | 300+ |
| High Quality + Form/Table | $15 | Forms and line-item tables | 300+ |
Source document highlighting, which maps an extracted value back to its position on the page, is an add-on at $0.01 per line coordinate lookup across all modes. For invoices, the $15 Form/Table mode is the one that keeps line-item columns aligned, which puts standalone parsing alone at 1.5 cents a page before any LLM call. LLMWhisperer returns layout-preserved text, not invoice fields, so it is the first half of an extraction pipeline rather than a finished one.
Yes. The core Unstract platform is published on GitHub under the AGPL-3.0 license and has more than 7,000 stars. Self-hosting removes the per-page fee, but you still pay for the LLM, the servers, a vector database and the engineering time to run and upgrade it. AGPL also requires you to publish your modifications if you offer a modified version to users over a network, which legal teams usually want to review before a production rollout.
The paid On-Prem Enterprise edition is a different thing: a supported, self-hosted deployment in your own AWS, Azure or GCP account with unlimited pages, SSO and the compliance attestations. It is quote only.
Unstract and InvoiceExtractor overlap on one job and differ on everything around it. This is an honest side-by-side for teams whose documents are mostly vendor invoices.
| Question | Unstract | InvoiceExtractor |
|---|---|---|
| Entry price | $499 a month for 5,000 pages | $49 a month for 2,500 pages |
| LLM cost | Your own provider bill after the trial | Included in the plan |
| Setup | Write and test prompts, deploy an API or pipeline | Upload an invoice, fields come back |
| Document types | Any: contracts, statements, forms, invoices | Invoices and bills only |
| Outputs | API, ETL to databases and warehouses | Excel, CSV, JSON, QuickBooks file, API |
| Self-hosting | Yes, AGPL edition or On-Prem Enterprise | No, cloud only |
| Where it wins | Many document types, pipelines, VPC deployment | Invoice-only teams that want results today |
If your roadmap includes contracts, loan packets and insurance forms flowing into a warehouse, Unstract is the more complete platform and the price reflects it. If the work is vendor invoices with line items going into a spreadsheet or your accounting system, a finished extractor gets you there without a prompt to write, at a fifth of the price per page or less with the model already paid for. You can test that claim on your own invoices in the uploader at the top of this page.
Comparing other developer-first options before you decide? The Reducto pricing breakdown and LlamaParse pricing per page cover the two platforms buyers most often shortlist next to Unstract, and the invoice data extraction API page shows the JSON our own endpoint returns.
Unstract Starter costs $499 a month for 5,000 pages and Growth costs $2,249 a month for 25,000 pages, with overage at $0.10 and $0.09 a page. Paid yearly they are $416 and $1,874 a month. Enterprise is quote only. Your LLM provider bills token usage separately after the 14-day trial.
Yes. Unstract offers a 14-day trial that includes LLMWhisperer, Azure OpenAI GPT-4o with 1M tokens, an embedding model and a Postgres vector database, so you can extract documents on day one. After the trial you choose a paid plan and connect your own LLM.
After the trial, yes. The pricing page says you connect your preferred stack once the trial ends, so you supply an OpenAI, Anthropic, Bedrock, Azure OpenAI or similar account, and that provider bills the tokens each extraction uses on top of the Unstract page fee.
LLMWhisperer costs $1 per 1,000 pages in Native Text mode, $5 in Low Cost, $10 in High Quality and $15 in High Quality with Form and Table, with 100 free pages a day. It is included at no extra charge inside paid Unstract plans.
Yes. The core Unstract platform is on GitHub under the AGPL-3.0 license. Self-hosting removes the page fee, but you still pay for the LLM, infrastructure and maintenance, and AGPL requires sharing modifications if you serve a modified version over a network.
Only on Enterprise. SOC 2 Type II, ISO 27001, HIPAA and GDPR with a DPA are listed for Cloud Enterprise and On-Prem Enterprise. Starter and Growth include encryption in transit and at rest and standard compliance, but not those attestations.
No, if the failure is on the Unstract side, such as a timeout or an unreadable file. Pages that process successfully are charged even if the extracted values are not perfect, because accuracy depends on your documents and prompts.
For vendor invoices exported to Excel, CSV, JSON or QuickBooks, InvoiceExtractor costs $49 a month for 2,500 pages and $149 for 10,000 pages, with the AI model included and no prompts to write. It handles invoices only, so Unstract remains the better fit for many document types.
Reducto pricing per 1,000 pages on the per-product rates that replaced credits.
LlamaParse pricing and LlamaCloud credit cost per page.
Extend AI pricing per credit and the real cost of an extracted page.
OpenAI OCR pricing per 1,000 invoice pages by model.
Invoice data extraction API that returns fields and line items as JSON.
Invoice line item extraction for quantities, unit prices and totals.
Start turning your invoices into clean, structured spreadsheet data.
USD
per month
per month
$288 charged today
Choose speed vs accuracy when extracting
| Base AI Faster | 2,500 pages |
| Pro AI Best accuracy | 500 pages |
Scale invoice extraction across your whole team with automation.
USD
per month
per month
$888 charged today
Choose speed vs accuracy when extracting
| Base AI Faster | 10,000 pages |
| Pro AI Best accuracy | 2,000 pages |
Enterprise‑grade invoice extraction, security, and controls.
USD
per month
per month
$ charged today
Choose speed vs accuracy when extracting
| Base AI Faster | pages |
| Pro AI Best accuracy | pages |
Secure, encrypted checkout
