Embedding Cost Calculator: Compare OpenAI, Cohere, Voyage, Google & Self-Hosted
The only unified multi-provider embedding calculator with vector database storage, batch API discounts, and a Year 1 total for every major provider. Enter your workload, see all options side by side.
Your Workload
~200,000 docs at 500 tokens each
$0.023/GB/month storage rate
Year 1 Cost by Provider (sorted cheapest first)
Assumptions: ~500 tokens/chunk; storage calculated from index size; query embedding at standard rate (no batch). Self-hosted estimated at A100 spot price. Verify all prices before committing. Optimization tips
How the Calculator Works
Cost formula
One-time embedding cost = (index tokens / 1,000,000) x rate. Monthly query cost = (queries/day x avg tokens x 30) / 1,000,000 x rate. Storage = (vector count x dimensions x 4 bytes) / GB x $/GB/month. Year 1 = one-time + monthly total x 12.
Vector count is estimated at one vector per 500-token chunk. Real chunk sizes vary by your pipeline.
Batch API behaviour
When batch mode is enabled, indexing tokens use the batch rate (50% off for OpenAI, 33% off for Voyage). Query embeddings always use the standard rate because queries are real-time and cannot be batched.
Self-hosted BGE-M3 cost is estimated at $1.50/hour A100 spot / 8,000 tokens/second throughput, giving approximately $0.052/M tokens at full GPU utilization.