Initial ingestion
One-time cost
RAG cost calculator
Budget estimate for a phased retrieval-augmented search system. All amounts in euros, excluding VAT.
Hybrid search
Per month, including cloud rental
Search with assistant
Per month, including hybrid search and cloud rental
Total first-year cost
First 12 months · search only
Phase 0 · Cloud rental
Monthly total
Server rental for the self-hosted hybrid search service, including vector and keyword index storage and backups.
Sizing hint from Phase 1 inputs
Enter the document count, pages and an embedding model in Phase 1 to estimate the index size.
Phase 1 · Initial ingestion
One-time total
One-time total = documents × pages/document × tokens/page ÷ (1 − overlap/100) × embedding rate ÷ 1,000,000 + additional compute
Used only for the index sizing hint.
Advanced ingestion assumptions
Affects the index size hint, not the embedding cost.
Phase 2 · Hybrid search
Monthly usage
Monthly usage = searches/month × (query embedding cost/search + reranking cost/search) + document update embedding cost/month
Monthly search usage breakdown
Query embeddings
New or updated documents
Reranking
Searches made directly, excluding assistant conversations, which are counted in Phase 3.
Each one is fully re-embedded with the Phase 1 assumptions.
Estimated searches per month: —
Phase 3 · Search with assistant
Monthly usage
Monthly usage = Phase 2 usage + assistant turns/month × [retrieval cost/turn + (input tokens/turn × blended input rate + output tokens/turn × output rate) ÷ 1,000,000]
Every turn runs one hybrid retrieval plus one generation.
Estimated assistant turns per month: —
Complete model input: system prompt, retrieved passages and conversation history.
Include billable reasoning tokens.
0 % is the conservative default. A stable system prompt and long conversations can reach 50 % or more; cache write premiums are ignored.
Model pricing sources · checked
| Model | Input / request rate | Cached input | Output | Unit |
|---|