Embedding API · billing diagnostic
Why is your embedding API bill high?
Direct answer
A embedding API bill almost always climbs from a metered unit crossing its free allowance — most often input tokens. Check the drivers below in order (biggest lever first), compare each against your plan's free tier in the vendor console, and look for a sudden usage change. These are the common causes, not a reading of your specific invoice.
The usual cost drivers — check in this order
Every provider in this category meters these. Past the free allowance, each is billed per unit — so a spike in any one moves the bill.
- ·input tokens — 100 = 100M tokens; embeddings bill on INPUT tokens only (no output). This is a metered unit: past the free allowance you pay per input tokens.
Tool-specific cliffs (from verified pricing risks)
Documented ways each tool's bill jumps — verify against your own plan and usage.
- ·Voyage 3 Lite: Smaller company than the hyperscalers — weigh continuity
- ·Google Gemini Embedding: 7.5x the price of Voyage 3 Lite at the same job — pay for the Google ecosystem/scale, not lower cost
- ·Google Gemini Embedding 2 (multimodal): $0.20/1M is 10x Voyage 3 Lite for text-only work
Where to check in your vendor console
- ·Open the usage/billing dashboard and find the line item that grew — it names the metered unit.
- ·Compare current usage to your plan's free allowance (the free-tier caps on that tool's pricing page).
- ·Look for a step change vs a gradual climb: a step usually means a plan change, a new feature, or crossing a free-tier cliff.
- ·Check egress/bandwidth and idle/always-on resources separately — they're common surprises the headline price hides.
These are the common causes for this category, not a diagnosis of your specific bill. Confirm in your vendor's usage dashboard, and contact vendor support if the numbers still don't add up. Prices and free tiers change — check the dated sources below.
FAQ
What usually makes a embedding API bill spike suddenly?
A sudden jump is most often a metered unit (input tokens) crossing the free-tier cap, a plan change, or a new feature — a step change rather than a gradual climb.
Can StackSays tell me my exact bill cause?
No — only your vendor's usage dashboard shows your actual usage. This page lists the likely causes and where to verify each; it never claims to read your invoice.
Official sources
- Voyage 3 Lite — Voyage AI pricing (official) · verified 2026-07-17
- Google Gemini Embedding — Google Gemini API pricing (official) · verified 2026-07-17