llmproviders.ai
← All providers

Amazon pricing

verified 2026-09-02

Amazon's Nova family on Bedrock — priced aggressively for high-volume enterprise workloads inside AWS.

data residency & retention
from Amazon's own policies · verified 2026-09-02
Headquarters
🇺🇸 United States (US)
Where inference runs
US, EU, Other
Inference runs in the AWS Region the customer selects (Bedrock spans US, EU, APAC and other commercial Regions), and content is encrypted and stored at rest in that Region.
Region pinning
Yes, self-serve
Region selection is inherent to AWS; cross-Region inference offers geography-bound profiles (US, EU, APAC) that keep routing within the geography.
Prompt retention
Not retained
Bedrock uses a zero-data-retention model by default (inputs and outputs are not stored); documented exceptions retain traffic up to 30 days for abuse review on specific third-party frontier models — not Amazon Nova.
Trains on API data
No
Customer content is not used to improve base models and is not shared with model providers.
Zero data retention
Available
ZDR is the default for most models including Nova; a configurable data-retention mode can enforce zero retention account- or project-wide.
current models · $ per 1M tokens
ModelInput /1MCached inOutput /1MContext
Nova 2 Lite
Nova
$0.30$0.07$2.501M
Nova 2 Pro
Nova
$1.25—$101M
Nova Premier
Nova
$2.50—$12.501M
Nova Lite
Nova
$0.06$0.01$0.24300K
Nova Micro
Nova
$0.04$0.0088$0.14128K
Nova Pro
Nova
$0.80$0.20$3.20300K
Amazon totals · 1M in + 1M out
Nova Micro
$0.1750
Nova Lite
$0.3000
Nova 2 Lite
$2.80
Nova Pro
$4.00
Nova 2 Pro
$11.25
Nova Premier
$15.00

Cost of 1M input + 1M output tokens. Bar length is square-root scaled so cheap models stay visible next to premium ones.