Cost formula
Ingestion cost = (pages × tokens per page ÷ chunks × embed price per token) + (optional LLM extract: pages × extract tokens / 1M × LLM price); Monthly refresh = new pages × same unit costs
Calculator
PDF & document ingestion cost calculator
Plan one-time (or re-index) cost for a document corpus: estimate tokens from pages, chunk with overlap for embeddings, and optionally price an LLM extract / OCR cleanup pass.
Document ingestion
Pipeline total
$0.0565total
500 docs · 6,000 pages · $0.000113 / doc
Extract / chunk prep$0.00
Embed$0.0565
Cost / document$0.000113
Cost / page$0.000009
Approx. chunks3,530
Cost / 1k pages$0.009412
Corpus
Pipeline inputs
Stages
Chunk → embed → store
API-priced stages only; vector host storage is outside this quote.
Raw corpus tokens2,400,000
Embed tokens (w/ overlap)2,823,480
Overlap inflation423,480
Total ingestion$0.0565
Pair with the RAG calculator for ongoing query cost after the index is stored.
Tokens-per-page is a planning estimate (layout, language, and OCR vary widely). Updated 2026-07-31. Approximate static list prices for planning only. Always verify against each provider’s official pricing page before production budgeting.
Guide
How to get value from this calculator
Estimate cost to ingest PDF and document corpora: page-to-token conversion, chunking, embedding, and optional LLM extraction passes. One-time ingestion spikes are often omitted from RAG budgets but dominate launch-month spend. Plan bulk upload events separately from steady-state query cost.
Why it matters
Document ingestion is a batch token event that can exceed a month of query cost in a single upload weekend. Teams that only budget query-time LLM calls miss the embedding bill for indexing thousands of pages.
How to use it
- Enter total pages or documents in the corpus to ingest.
- Set estimated tokens per page after parsing and average chunk size.
- Choose embedding model and optional LLM extraction pass for structured fields.
- Review pipeline stage costs for parse, chunk, embed, and optional LLM extract.
- Add re-embedding cost if you change chunk strategy or embedding model.
- Combine with the RAG query calculator for full pipeline monthly spend.
Planning tips
- Parse and clean PDFs locally to avoid sending raw binary through token-heavy paths.
- Skip LLM extraction when simple text extraction suffices for your retrieval quality bar.
- If one pipeline stage dominates cost, cut that stage before shrinking the whole corpus.
- Incremental ingestion of new docs only—avoid full re-embed on unchanged files.
- Large tables and scanned images inflate tokens per page—sample real documents before estimating.
- Batch embedding APIs reduce cost for bulk ingestion when your provider supports them.
Frequently asked questions
Does this include storage costs?
No. This covers parsing, embedding, and optional LLM extraction API costs. Vector store and object storage are separate.
How accurate is tokens-per-page estimate?
Varies widely by document type. Use the token counter on representative pages for better accuracy than generic defaults.
Are embedding prices official?
CentsPerToken uses approximate list prices for planning. Verify embedding rates on your provider's pricing page.
Should I re-ingest the full corpus when switching embed models?
Yes. New embedding models require re-indexing all chunks—a significant one-time cost to model before migrating.