LlamaParse Pricing 2026: Cost Per Page, LlamaParse API Pricing and Credit Costs

LlamaParse pricing at a glance

LlamaParse does not publish a price per page. It publishes a price per credit, and every operation you run burns a different number of credits. One credit is worth $0.00125, because LlamaIndex sells credits at $1.25 per 1,000 in both its North America and Europe regions. Everything on this page is that one rate multiplied by the credit costs LlamaIndex documents.

Last updated August 2026. Plan prices and credit rates were re-fetched from the official LlamaIndex pricing page and the LlamaParse developer documentation on the date shown. Confirm current figures before you commit budget, since model pricing moves.

LlamaCloud plans and included credits

Four tiers, all of which include access to every module (Parse, Extract, Classify, Split, Index). What changes between them is volume, concurrency, support, and deployment options.

Read the fourth column carefully, because it is the part most pricing write-ups skip. Starter costs $50 and hands you exactly $50 worth of credits. Pro costs $500 and hands you exactly $500 worth of credits. The paid plans carry no volume discount at all. You pay $1.25 per 1,000 credits whether you spend $50 or $500 a month, and the subscription fee buys concurrency, support, and deployment options rather than cheaper pages. That is unusual in this market, where per-page rates normally fall as commitment rises.

LlamaParse cost per page by parse tier

The v2 API exposes four parse tiers. LlamaIndex picks the underlying model for each tier, so you choose a quality and cost band rather than a specific model.

The gap between the cheapest and dearest tier is 45x, which is why a single quoted "LlamaParse price" is close to meaningless without naming the tier. If you have seen the figure $0.10 a page quoted for LlamaParse, it is wrong: the most expensive documented tier is $0.05625.

Which tier do invoices actually need?

This is where budgets go wrong. Invoices are table documents. They carry line items, multi-column totals blocks, and frequently arrive as scans or phone photos. LlamaIndex own guidance points that shape of document at the Agentic tier, describing it as the default recommendation for most real-world PDFs with tables, scanned pages, multi-column layouts, charts and diagrams.

So the honest planning number for invoice work is 10 credits a page, or $12.50 per 1,000 pages, not the 1 credit headline. Budget on the Fast tier and you will be out by a factor of ten.

Parse is not extraction, and that changes the bill

A parse returns the document as clean markdown. It does not return a vendor name, an invoice number, a due date, a tax total, or a line-item table mapped to fields. Getting typed fields means calling Extract, and Extract bills the extract tier plus the parse tier underneath it.

Moving from text on the page to invoice fields in a row multiplies the LlamaParse bill by 2.5x, from $12.50 to $31.25 per 1,000 pages at the default Agentic settings. That step is the one most cost models leave out, and it is the number to plan against if the goal is structured invoice data rather than markdown for a retrieval index.

Classify and Index credit costs

Two other modules bill separately if your pipeline uses them.

The chat turn rate is worth flagging. At 100 credits a turn, a chat interface over your documents costs 12.5 cents per exchange, which is 10 times what it costs to parse a page at the Agentic tier. A support tool answering a few thousand questions a month will spend more on turns than on parsing.

How many pages each plan actually covers

Included credits divided by credits per page. These are parse-only figures, so halve them again roughly if you are running Extract.

A finance team processing 2,000 invoices a month with structured extraction burns 50,000 credits, which overshoots Starter and lands in pay-as-you-go territory at $1.25 per 1,000 credits for the excess, or roughly $62.50 a month all in. On the Free plan, requests that would consume credits you do not have return a 402 error, and the allowance resets at the start of the next billing cycle.

LlamaParse compared with the other options for invoice data

Per 1,000 invoice pages, using each vendor published rates and the tier each one recommends for table-heavy documents. Figures for the other vendors were verified against their own documentation on the dates noted in our individual pricing pages.

The comparison that matters: LlamaParse Extract at $31.25 per 1,000 pages costs more than three times AWS Textract AnalyzeExpense or the Azure prebuilt invoice model at $10.00, and both of those return typed invoice fields without you writing a schema. That is not a knock on LlamaParse so much as a statement of what it is for. It is a general-purpose document parser built for retrieval pipelines across dozens of formats, not an invoice product, and it is genuinely strong at that job.

Where LlamaParse is the right call

If you are building a retrieval system over mixed documents and invoices are one input among many, LlamaParse earns its rate. It handles a wide format range, plugs straight into LlamaIndex, and the Fast tier at $1.25 per 1,000 pages is very cheap for clean text. If invoices are the whole job and you want vendor, date, totals and line items in a spreadsheet without building and maintaining a schema, a per-page parser is the wrong shape of tool and the cheaper answer is a flat-rate invoice product.

Frequently asked questions about LlamaParse pricing

How much does LlamaParse cost?

LlamaParse costs $1.25 per 1,000 credits, and pages consume 1 to 45 credits depending on the parse tier. That works out to $1.25 per 1,000 pages on the Fast tier and $56.25 on Agentic Plus. Paid plans start at $50 a month for the Starter tier with 40,000 credits included.

Is LlamaParse free?

LlamaParse has a free plan with 10,000 credits a month, which covers 10,000 pages on the Fast tier or 1,000 pages on the Agentic tier. It is not free beyond that allowance. Once credits run out, credit-consuming API requests return a 402 error until the next billing cycle resets them.

Is LlamaParse open source?

No. LlamaParse is a commercial hosted service from LlamaIndex and requires an API key. The LlamaIndex framework itself is open source, which causes the confusion. If you need a genuinely open-source parser you can self-host, Docling is MIT-licensed and free for commercial use.

What is LlamaParse used for?

LlamaParse converts documents into clean, structured markdown for AI and retrieval pipelines. It handles PDFs, presentations, spreadsheets and images, preserving tables and reading order so a language model can use the output. It is most commonly used as the ingestion step in a RAG system built on LlamaIndex.

How many credits does an invoice page use?

Plan on 10 credits per page for invoice parsing, the Agentic tier, because invoices carry tables, multi-column totals and frequently arrive as scans. If you need typed fields rather than markdown, Agentic Extract bills 25 credits per page, combining 15 extract credits with 10 parse credits underneath.

Does LlamaParse get cheaper at higher volume?

No. The credit rate is $1.25 per 1,000 on every paid tier, and included credits are worth exactly what the plan costs: Starter is $50 for $50 of credits, Pro is $500 for $500 of credits. Higher plans buy concurrency, support and deployment options rather than a lower per-page rate.

What is the difference between LlamaParse and LlamaExtract?

Parse returns the document as markdown. Extract returns typed fields against a schema you define. Extract bills both operations together, so the default Agentic configuration costs 25 credits a page rather than 10. If you want structured invoice data, Extract is the one you need and the one to budget for.