Amazon API Cost Calculator

How much does the Amazon API cost per month?

For Amazon Nova Micro at 200K requests/month, 2,400 input tokens and 350 output tokens per request costs about $12.12 per month after 30% cache use and 100% batch share. Across Amazon's 3 priced models, the cheapest default ranking is Amazon Nova Micro at $24.25 per month.

Verified 2026-06-21

Pricing data as of June 2026. Sources: Amazon pricing and model documentation. Parameters are shareable in this URL.

Amazon Nova Micro: estimated monthly cost

$12.12

Formula: provider-specific cached input multiplier × cache rate, plus uncached input and verbosity-adjusted output; cache writes are amortised over the provider TTL, then batch savings are applied.

Per request$0.000
Per day$0.404
Per month$12.12
Per year$145

OpenAI scenario sensitivity

The default is a server-rendered estimate. Change the cache and batch shares to see when a cheaper qualified model overtakes the selected model; the URL is shareable and preserves the inputs.

Default$12.12/month
No caching or batch$24.25/month
50% cache, 50% batch$18.19/month
100% cache, 100% batch$12.12/month

Ranked cost — 200K requests/month

ModelProviderList monthlyEffective monthlyVerbosityRank Δ
Amazon Nova MicrobudgetAmazon$26.60$12.120.76×
Amazon Nova LitebudgetAmazon$45.60$22.130.92×
Amazon Nova PromidAmazon$608$304

Levers live on Amazon

Batch API: up to 50%Model verbosity: up to 94%Context trimming: up to 47%

The billing gotcha

Amazon Nova is billed through AWS Bedrock rather than a standalone Nova account: the token estimate below covers model input and output charges, while your real invoice can also include regional AWS data-transfer, guardrail, knowledge-base, or orchestration costs. Nova's batch discount applies to eligible asynchronous Bedrock jobs, so selecting a batch share is only realistic when your application can submit and collect work later instead of waiting on Converse synchronously.

Related

Amazon provider profile →Compare against another provider →Global cost calculator →How to reduce LLM API costs →