RAG Cost Calculator
Price a RAG pipeline end-to-end: queries, retrieved tokens and LLM prices.
Queries per day
Retrieved tokens / query
Answer tokens / query
Blended $/1M tokens
Per-query = tokens/1M × blended $/1M + $0.0005 retrieval. Monthly = per-query × q/day × 30.
Live estimate
- Tokens / query
- 1,900
- Cost per query
- $0.00
- Daily cost
- $4.80
- Monthly cost
- $144.00
- Annual cost
- $1,752.00
RAG benchmarks
Retrieval tokens usually exceed answer tokens 2–4× — top-k is the cost lever.
| top-k | Tokens/query | $/1K queries |
|---|---|---|
| top-3 | ~1,200 | $1.70 |
| top-5 | ~1,900 | $2.40 |
| top-10 | ~3,500 | $4.00 |
Frequently asked questions
- How is rag cost calculated?
- Per-query = tokens/1M × blended $/1M + $0.0005 retrieval. Monthly = per-query × q/day × 30. Retrieval tokens usually exceed answer tokens 2–4× — top-k is the cost lever.
- What inputs do I need for the rag cost calculator?
- Enter queries per day, retrieved tokens / query, answer tokens / query, blended $/1m tokens. Defaults reflect typical 2026 usage; replace them with your own vendor quote or last month of logs for an exact figure.
- What is a typical rag cost benchmark?
- 2,000 queries/day at ~1,900 tokens and $1/1M costs ~$144/month.
- RAG vs fine-tuning cost?
- RAG is pay-as-you-go and updateable; fine-tuning is upfront with cheaper prompts — many teams use both.
- How can I lower my rag cost?
- Cut volume waste first (shorter durations, fewer retries, tighter prompts), then route to cheaper tiers for drafts, enable caching/batch where available, and cap per-user or per-job spend with alerts. Re-run the calculator after each change to size the saving.
- Is this calculator free?
- Free forever, no sign-up, runs 100% in your browser. Copy results or export CSV for your budget doc.