unitcostai

Retrieval Cost

How much does retrieval add when I change chunk size or depth?

How this tool works

This simulator runs the deterministic retrieval-workflow economics model twice with shared assumptions and isolates what chunk count and chunk size changes do to cost per user and break-even price.

How It Works

  1. Set provider/model plus workload assumptions used in both runs.
  2. Set baseline and candidate chunk assumptions.
  3. Compare retrieval tokens, cost deltas, and break-even deltas.

Formula

retrieval_tokens_per_request = retrieved_chunks * tokens_per_chunk

total_cost_delta = cost_candidate - cost_baseline

Assumptions and Units

  • Currency: USD
  • Token unit: token
  • Baseline and candidate use the same non-chunk assumptions
  • Pricing source: daily pricing snapshot in repo, no runtime scraping

Related resources: Rerank Cost, Cache Savings, Prompt Overhead, RAG or Long Prompt, Indexing Cost, AI Docs Assistant Cost per User, How To Choose Chunk Size and Chunk Count, What Is RAG?, RAG Cost Components Explained.

Pricing snapshot: 2026-07-19Provider: OpenAIModel: GPT-5 Mini

Step 1 Provider and Model

iChoose the model row used for both baseline and candidate retrieval assumptions.

Step 2 Quick Mode

iSet the baseline scenario first, then tune chunk assumptions.

Evaluate chunk compression before changing reranking or model tier.

Optional Advanced assumptions

iAdjust baseline retrieval and advanced cost assumptions only after Quick Mode is close.
Show advanced inputs

Scenario actions

Copy scenario URL

Paste into ChatGPT or Claude, or share with a teammate.

Save and track this scenario

Track pricing drift on this scenario and get an email if the latest result changes.

How tracking works

After you click Save and track, we carry this exact calculator state into the tracked-scenarios page so you can sign in and confirm the save.

We save your assumptions and the pricing snapshot used for this result.

When a newer pricing snapshot lands, we recompute the same scenario, show what changed, and email you if the latest result moved.

1 tracked scenario free, then $12/mo or $120/yr for up to 25 tracked scenarios.

Headline metric

Candidate chunk plan lowers cost

Total cost delta per user / month: $-0.0081

Candidate retrieval tokens / request: 1,320 vs baseline 1,680.

Cost delta / user / month

$-0.0081

Retrieval token delta / request

-360

Break-even delta

$-0.0081

Monthly cost delta

$-5.26

Totals

iBaseline vs candidate totals under the same retrieval assumptions.
Cost per request
Baseline
$0.01634
Candidate
$0.01627
Delta
-$0.00007
Cost per user/month
Baseline
$1.9607
Candidate
$1.9526
Delta
-$0.0081
Gross margin %
Baseline
96.0%
Candidate
96.0%
Delta
+0.0%
Break-even price
Baseline
$1.9607
Candidate
$1.9526
Delta
-$0.0081
MetricBaselineCandidateDelta
Cost per request$0.01634$0.01627-$0.00007
Cost per user/month$1.9607$1.9526-$0.0081
Gross margin %96.0%96.0%+0.0%
Break-even price$1.9607$1.9526-$0.0081

Component Breakdown

iBaseline and candidate components are computed independently, then differenced.
GenerationiModel input/output token spend for requests.
Baseline
$0.114
Candidate
$0.114
Delta
$0
RetrievaliExtra model input spend from retrieved context chunks.
Baseline
$0.0504
Candidate
$0.0396
Delta
-$0.0108
RerankingiReranker cost based on docs scored per request.
Baseline
$2.4
Candidate
$2.4
Delta
$0
Embeddings IngestioniAmortized per-user share of the fixed monthly corpus embedding refresh cost.
Baseline
$0
Candidate
$0
Delta
$0
Vector DbiVector database query cost across all requests.
Baseline
$0.0018
Candidate
$0.0018
Delta
$0
CacheiSavings from cache hits. Negative means lower total cost.
Baseline
$-0.6536
Candidate
$-0.6508
Delta
+$0.0027
InfraiNon-model infra overhead per request.
Baseline
$0.048
Candidate
$0.048
Delta
$0
ComponentBaselineCandidateDelta
GenerationiModel input/output token spend for requests.$0.114$0.114$0
RetrievaliExtra model input spend from retrieved context chunks.$0.0504$0.0396-$0.0108
RerankingiReranker cost based on docs scored per request.$2.4$2.4$0
Embeddings IngestioniAmortized per-user share of the fixed monthly corpus embedding refresh cost.$0$0$0
Vector DbiVector database query cost across all requests.$0.0018$0.0018$0
CacheiSavings from cache hits. Negative means lower total cost.$-0.6536$-0.6508+$0.0027
InfraiNon-model infra overhead per request.$0.048$0.048$0
Sensitivity RankingiDelta in total cost if one variable increases by 10%.
VariableCost delta %
Requests Per User MonthiUser activity level per month.10.00%
Rerank DocsiDocs reranked per request.9.22%
Cache Hit RateiFraction of requests served by cache.-3.33%
Output TokensiGenerated tokens per request.0.32%
Retrieved ChunksiRetrieved chunk count per request.0.15%
Tokens Per ChunkiAverage chunk size in tokens.0.15%
Input TokensiPrompt-side tokens per request.0.12%
Vector Queries Per RequestiVector query count per request.0.01%
Monthly Active UsersiActive-user estimate used to amortize fixed monthly embedding refresh.-0.00%

Assumptions and Units

iExplicit assumptions keep this comparison reproducible.
  • CurrencyUSD
  • Token unittoken
  • Pricing snapshot2026-07-19
  • Selected model rowOpenAI/GPT-5 Mini
  • Comparison ruleOnly retrieval chunk assumptions change; non-retrieval inputs stay shared
  • Volume basisFixed monthly terms and business totals use monthly active users as the denominator

Recommended Next Step

iUse this section to translate chunk deltas into the next quality and infrastructure checks.

Validate infra and retrieval assumptions, then confirm quality on sampled traffic.

Chunking Playbook

How To Choose Chunk Size and Chunk CountRAG Cost Components Explained

Compare infra providers

View Infra Recommendations

Sources and Snapshot

iPricing comes from the current dated snapshot.

Active Pricing Row

Candidate

OpenAI / GPT-5 Mini

  • Input tokens$0.25 / 1M
  • Output tokens$2 / 1M

Shared retrieval defaults

  • Embedding input$0.02 / 1M
  • Rerank docs$1 / 1K

Continue Analysis

Switch tools

Read guides