llmproviders.ai
← All providers

Microsoft Azure pricing

verified 2026-09-02

OpenAI's models served on Azure AI Foundry — same rates as OpenAI direct, with Azure networking, compliance, and billing.

data residency & retention
from Microsoft Azure's own policies · verified 2026-09-02
Headquarters
🇺🇸 United States (US)
Where inference runs
US, EU, Other
Standard deployments process prompts within the customer-specified Azure geography (US, EU, and many others); DataZone deployments pin to the US or EU member states; Global deployments may process anywhere the model is deployed.
Region pinning
Yes, self-serve
Customers self-select region/geography at deployment time; data at rest, including the abuse-monitoring store, stays in the customer-designated geography even for Global deployments.
Prompt retention
Retained · up to 30 days (flagged content)
Models are stateless, but flagged prompts/completions may be stored ~30 days in a geography-local abuse-monitoring store for review, unless the customer is approved for modified abuse monitoring.
Trains on API data
No
Prompts and completions are not used to train foundation models without explicit permission, and are not available to OpenAI.
Zero data retention
Available
Customers meeting Limited Access eligibility can apply for modified abuse monitoring, which turns off prompt storage and human review; automated real-time review still occurs without storage.
current models · $ per 1M tokens
ModelInput /1MCached inOutput /1MContext
GPT-5.6 Sol
GPT-5
$5$0.50$301.1M
GPT-5.6 Terra
GPT-5
$2$0.20$121.1M
GPT-5.6 Luna
GPT-5
$0.20$0.02$1.201.1M
GPT-5.5
GPT-5
$5$0.50$301.1M
Microsoft Azure totals · 1M in + 1M out
GPT-5.6 Luna
$1.40
GPT-5.6 Terra
$14.00
GPT-5.6 Sol
$35.00
GPT-5.5
$35.00

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