Azure AI Foundry cost calculator

Estimate your monthly Microsoft Foundry bill.

See what it costs to run Microsoft Foundry AI models and the MultiModel AI portal in your own Azure tenant. Set your seat count and user mix, and get token and hosting costs in seconds.

What the estimate covers
  • Model token costs for three usage tiers across four model levels, per user per month
  • AI portal hosting on Azure, scaled to your seat count
  • Any seat count from 20 to 1,500, in 20-seat steps
  • Typical office-worker usage, excluding software development
  • Monthly figures in USD, rounded
Price estimate

Estimate the cost of your AI usage.

Set your seats, the models you'll offer, and how your people will use them to estimate your monthly Azure bill.

1

Seats

How many employees will have access. Drives both token costs and hosting.

100 seats
20in 20-seat increments1,500
2

Models

You choose which users can access each model. For this calculator, all weekly and power users will favor more powerful models when available.

3

Usage

Start from a preset, then drag the handles to set your user mix.

Entry level 65% Weekly users 25% Power users 10%

Monthly cost breakdown

Entry level users
65 users × $0.54
$35
Weekly users
25 users × $1.93
$48
Power users
10 users × $16.76
$168
AI portal hosting
scales with 100 seats
$353
Total monthly
100 seats
$604
Average cost per user
total ÷ 100 seats
$6.04

Estimated annual cost

Total annual
$604 / month × 12
$7,248
Annual cost per user
annual ÷ 100 seats
$72
Cost details
Entry level users
30 interactions / mo
65 users$0.54
/ user / mo
Weekly users
100 interactions / mo
25 users$1.93
/ user / mo
Power users
500 interactions / mo
10 users$16.76
/ user / mo
All users
Blended average token cost
100 users$2.51
/ user / mo

Assumptions

Share of each tier's interactions by model level, with every level on
User tierInteractions / moLiteStandardReasoningFrontier
Entry level3010%70%20%–
Weekly1005%70%20%5%
Power50010%20%50%20%
  • Turning a model level off shares its use across the levels that are still on.
  • Compared with Standard, an interaction costs 0.05 times as much on Lite, 4 times on Reasoning (2 times the per-token price, 2 times the tokens), and 20 times on Frontier (5 times the price, 4 times the tokens). Token volumes are estimates.
  • On Standard models alone, entry level, weekly, and power users cost $0.36, $1.22, and $6.08 a month.
  • AI portal hosting is $340 a month at 20 seats, $420 at 500, and about $587 at 1,500.
  • Figures are rounded. Currency is USD.
How the estimate works

Three inputs, one monthly figure.

The calculator multiplies users in each tier by that tier's monthly token cost, then adds the cost of hosting the AI portal on Azure.

01

Set your seats

Choose how many of your employees will have access, from 20 to 1,500 in 20-seat steps. Seat count drives both token costs and hosting.

02

Split users into tiers

Not everyone uses AI the same amount. Divide your seats between entry level, weekly, and power users, each with its own monthly cost.

03

Add portal hosting

Hosting the MultiModel portal in your Azure tenant adds a monthly cost that rises gently with seats, from about $340 to about $587.

Assumptions

What sits behind the numbers.

The figures are planning estimates for typical office work. Your actual Azure bill depends on the models you enable and how your staff use them.

Three usage tiers

Entry level users run about 30 interactions a month, weekly users about 100, and power users about 500. On Standard models alone, that costs $0.36, $1.22, and $6.08 per user.

Hosting that scales with seats

Portal hosting is estimated at $340 a month for 20 seats, rising in a straight line to $420 a month at 500 seats and about $587 a month at 1,500 seats.

Blended and per-user cost

The blended average is token cost divided by seats. The average cost per user is the full monthly total, hosting included, divided by seats.

Office work, not development

Tiers reflect typical office usage and exclude software development. Contact us for an estimate on a development project.

Blended model mix

Tier prices are blended across models, not priced per model. Based on publicly available Azure pricing.

Azure costs only

The estimate covers Azure tokens and hosting. MultiModel platform fees are separate: see pricing.

FAQ

Azure OpenAI and Foundry costs: FAQ

What to know when you budget for running AI models in your own Azure tenant.

Talk to us
What does this calculator include?

Two things: the model token costs for your users, estimated per user per month for three usage tiers, and the cost of hosting the MultiModel AI portal on Azure. Token costs scale with your seat count and user mix. Hosting scales with seat count, from about $340 a month at 20 seats to about $420 a month at 500 seats and about $587 a month at 1,500 seats.

Why would our Azure AI costs appear on our own Azure bill?

Because the portal and your models run in your own Azure tenant. That is the point of the deployment: your prompts and member data stay inside a cloud boundary you control, under your existing Microsoft agreement. See AI on your network for deployment in Azure, AWS, or Google Cloud.

Does the estimate include MultiModel's own fees?

No. The calculator estimates Azure costs only: model tokens and portal hosting. For the MultiModel platform itself, see pricing or contact us.

How should we choose the user mix?

Start with the Typical preset (65% entry level, 25% weekly, 10% power users) and adjust. If you are rolling AI out to every branch and back-office employee, expect your mix to look more like Light (75/20/5) in the first months. If you are starting with a smaller group of heavy users, such as your compliance, lending, or marketing teams, Heavy (25/50/25) is closer.

Why does the estimate leave out software development?

The usage tiers are based on typical office work: drafting, summarizing, research, and analysis. Software development can use far more tokens per person and varies a lot by project, so it doesn't fit a per-seat average. Contact us and we'll estimate development usage for your specific project.

Does running models in our Azure tenant keep member data private?

It keeps data inside the perimeter you define, and MultiModel adds controls on top. Account numbers, SSNs, and other personal data are redacted on your employee's device before a prompt is sent, prompts are excluded from model-provider training, including OpenAI and Microsoft, and every prompt is logged for audit. See security & compliance and data loss prevention.

Which models should we budget for?

The calculator prices four model levels: Frontier (such as Claude Fable), Reasoning (such as Claude Opus), Standard (such as Claude Sonnet), and Lite (such as Claude Haiku). Per token, Frontier costs about 5 times Standard, Reasoning about 2 times, and Lite about a twentieth. Reasoning and Frontier models also use more tokens per interaction, because they think before answering, write longer answers, and get the heavier work, so the calculator assumes Reasoning uses 2 times the tokens of Standard and Frontier 4 times. In all, a Reasoning interaction costs about 4 times a Standard one and a Frontier interaction about 20 times. Turn levels on or off to match the models you plan to offer, and the calculator shares each tier's use across them. Treat the result as a planning figure: actual spend depends on which models your staff choose for which work. Our recommended AI models guide shows which model we suggest for which kind of banking work.

Want an estimate built on your own usage?

Tell us your seat count, departments, and the models you plan to enable. We'll size the Azure costs with you, including development workloads the calculator leaves out.