AI Cost Calculators
AI Embedding Cost Comparison Calculator
Compare current OpenAI, Google Gemini, Mistral, Voyage AI, and custom embedding costs across initial indexing, refreshes, queries, and first-year usage.
Enter Your Embedding Workload
Compare initial indexing and continuing monthly embedding usage.
Embedding usage used for comparison
Initial document tokens: 92,000,000
Monthly refresh tokens: 9,200,000
Monthly query tokens: 10,000,000
Monthly recurring tokens: 19,200,000
First-year embedded tokens: 322,400,000
Search queries: 500,000 per month
Embedding Cost Comparison
The selected model is shown first, followed by a paid list-price comparison for the same workload.
Selected model first-year cost
Initial indexing
$1.84
Monthly recurring
$0.384
Per 1M documents
$18.40
Lowest cost · OpenAI
text-embedding-3-small
Paid list price
Per 1M tokens
$0.02
Initial
$1.84
Monthly
$0.384
First year
$6.448
Voyage AI
voyage-4-lite
Paid usage after any free allowance
Per 1M tokens
$0.02
Initial
$1.84
Monthly
$0.384
First year
$6.448
Voyage AI
voyage-4
Paid usage after any free allowance
Per 1M tokens
$0.06
Initial
$5.52
Monthly
$1.152
First year
$19.344
gemini-embedding-001 · Batch
Text-only batch paid tier
Per 1M tokens
$0.075
Initial
$6.90
Monthly
$1.44
First year
$24.18
gemini-embedding-2 · Text · Batch
Multimodal model using text input
Per 1M tokens
$0.10
Initial
$9.20
Monthly
$1.92
First year
$32.24
Mistral
mistral-embed
Paid list price
Per 1M tokens
$0.10
Initial
$9.20
Monthly
$1.92
First year
$32.24
Voyage AI
voyage-4-large
Paid usage after any free allowance
Per 1M tokens
$0.12
Initial
$11.04
Monthly
$2.304
First year
$38.688
OpenAI
text-embedding-3-large
Paid list price
Per 1M tokens
$0.13
Initial
$11.96
Monthly
$2.496
First year
$41.912
gemini-embedding-001 · Standard
Text-only standard paid tier
Per 1M tokens
$0.15
Initial
$13.80
Monthly
$2.88
First year
$48.36
Voyage AI
voyage-context-3
Contextualized chunk embeddings
Per 1M tokens
$0.18
Initial
$16.56
Monthly
$3.456
First year
$58.032
Voyage AI
voyage-code-3
Code retrieval embeddings
Per 1M tokens
$0.18
Initial
$16.56
Monthly
$3.456
First year
$58.032
gemini-embedding-2 · Text · Standard
Multimodal model using text input
Per 1M tokens
$0.20
Initial
$18.40
Monthly
$3.84
First year
$64.48
Selected model: OpenAI — text-embedding-3-small
Lowest paid list-price option: OpenAI — text-embedding-3-small
Possible first-year saving against selected model: $0.00
Possible monthly saving against selected model: $0.00
Budget status: $99.616 remaining
* Important: Calculated from the values currently shown. Default usage values are examples, so change them to match your expected usage. Built-in OpenAI, Google Gemini, Mistral, and Voyage AI embedding rates were checked on June 19, 2026. Final charges may include free-tier eligibility, promotional credits, vector storage, vector reads and writes, reranking, LLM generation, subscriptions, taxes, discounts, and implementation work.
Embedding cost is not limited to the first document upload. Content refreshes, chunk overlap, and search-query embeddings can create continuing usage. This calculator compares the full embedding lifecycle across current paid provider rates.
Comparing the Full Embedding Lifecycle
Enter the number of source documents, their average token size, and the overlap added during chunking. The calculator estimates the initial token volume needed to create the document embeddings.
Add the percentage of content refreshed each month, monthly search queries, and average tokens in each query. The result separates initial indexing cost from recurring monthly embedding cost.
The comparison also shows a first-year total, the selected model's cost, the lowest paid list-price option, and the possible saving for the same workload.
Why Initial and Monthly Costs Are Different
Initial indexing embeds the whole knowledge base. Monthly operating usage usually includes only changed documents and new search queries.
A large existing document collection may therefore create a noticeable one-time indexing cost, while a stable knowledge base can have a much smaller monthly refresh cost.
A fast-changing product catalogue, news archive, support centre, or code repository may need a higher refresh percentage.
Using Chunk Overlap Carefully
Chunk overlap can help preserve context between neighbouring passages, but every repeated token increases embedding usage. A 15% overlap means the calculator adds 15% to the document token volume.
This is a planning estimate. Real token use depends on the tokenizer, document cleaning, chunking method, metadata, and whether titles or summaries are embedded with each chunk.
Practical Decisions This Tool Supports
- Choose an embedding API before building a RAG system.
- Estimate the first document-indexing bill.
- Plan monthly re-embedding and search-query costs.
- Compare standard and batch embedding prices.
- See the effect of chunk overlap on token usage.
- Compare a public price with a private provider quote.
- Estimate first-year embedding spend before scaling.
Costs Not Included in This Comparison
The result covers embedding API usage only. Vector database storage, vector reads and writes, reranking, LLM generation, hosting, data transfer, monitoring, taxes, subscriptions, and implementation work may add separate costs.
Use the RAG Cost Calculator when planning the full retrieval stack rather than only the embedding layer.
Official Pricing Sources
Built-in paid prices were checked on June 19, 2026 against official OpenAI, Google Gemini API, Mistral, and Voyage AI documentation. Only current listed models used by this calculator are included.
Frequently Asked Questions
What usage does the embedding comparison include?
It includes the first embedding pass for the document collection, repeated tokens caused by chunk overlap, monthly content refreshes, and embeddings created for search queries.
Why does chunk overlap increase embedding cost?
Overlap repeats part of the source text in neighbouring chunks. The repeated text is embedded again, so it increases the total token volume.
Are free tiers included?
No. The built-in comparison uses paid list prices so providers can be compared on the same recurring basis. Free allowances, credits, promotions, and negotiated discounts may reduce the actual bill.
Does a lower embedding API price always mean a lower total RAG cost?
No. Vector dimensions, storage, retrieval quality, language support, latency, batch support, and the number of chunks can affect the total system cost and usefulness.
Why are Google standard and batch options separate?
Google publishes separate paid rates for standard and batch embedding requests. Choose the option that matches how the workload will be processed.
Can I compare another embedding provider?
Yes. Add a custom provider and enter its current price per one million tokens.