Model your Copilot Credit spend by light, medium, and heavy task volume, apply the prepay discount tiers, and compare the same work on Claude direct at per token rates. Everything updates live.
Pre-filled with a 200 user example. Change any field and the cost updates instantly. Hover the small print for the official Microsoft ranges.
Rates per million tokens: Sonnet 4.6 $3 in / $15 out, Opus 4.8 $5 / $25, Haiku 4.5 $1 / $5. Anthropic list, June 2026.
Microsoft Copilot Credits, annual
$0
at list, pay as you go
| Tier | Tasks / mo | MS cost | Claude |
|---|---|---|---|
| Light | 0 | $0 | $0 |
| Medium | 0 | $0 | $0 |
| Heavy | 0 | $0 | $0 |
Planning estimate on Microsoft published ranges and Anthropic list rates, June 2026. Not a quote. The Claude figure is raw inference; it excludes the runtime, retrieval, identity, and governance that credits bundle.
Start with your real deployment, not a vendor forecast. Set the user count, then estimate how many light, medium, and heavy tasks each user runs in a month. The heavy count matters most, because a heavy task costs more than seven times a light one.
The calculator is built on Microsoft's published task ranges and Anthropic's list token rates. Both are documented and current to June 2026. The token counts behind each tier are what map a Copilot task to a Claude cost.
| Tier | Credits | Default tokens in / out | Cost at one cent |
|---|---|---|---|
| Light | 70 to 200 | 25,000 / 1,000 | $0.70 to $2.00 |
| Medium | 400 to 600 | 80,000 / 4,000 | $4.00 to $6.00 |
| Heavy | 1,500 plus | 600,000 / 12,000 | Over $15.00 |
Pay as you go is one cent per credit. The Pre-Purchase Plan shaves 5 to 20 percent off that, but only the deepest tier reaches 20 percent, and unused prepaid credits expire at term end. The calculator applies the discount you pick to the credit price, so you can see the effective per credit cost fall toward the eight tenths of a cent floor.
Enter your user count and the number of light, medium, and heavy tasks each user runs per month. The calculator multiplies your task volume by the credits per task, applies the credit price and any prepay discount, and shows the monthly and annual cost. It also prices the same token volume on Claude direct so you can see the build versus buy gap.
Microsoft prices by task complexity. A light Cowork task runs 70 to 200 credits, a medium task 400 to 600, and a heavy task over 1,500, at one cent per credit. The calculator pre-fills the midpoints and lets you adjust each, because your real distribution sets the bill.
Each task tier maps to an input and output token count. The calculator applies Anthropic list rates, Sonnet 4.6 at 3 dollars in and 15 out, Opus 4.8 at 5 and 25, Haiku 4.5 at 1 and 5 per million tokens, then optionally halves the cost for batch and reduces input cost for prompt caching. The result is the raw inference cost, not the managed agent.
Yes. Pay as you go is one cent per credit. The Pre-Purchase Plan discounts that 5 to 20 percent depending on commitment size, so the effective floor is eight tenths of a cent at the deepest tier. Select a tier to see the discounted cost, but remember unused prepaid credits expire.
No. It is a planning estimate built on Microsoft published ranges and Anthropic list rates as of June 2026. Your actual credits per task vary with how much context each task pulls. Use it to size a budget and a negotiation position, then validate against real pay as you go consumption.
The per task costs, the prepay floor, the Microsoft versus Claude comparison, and six moves to set before you provision. Reads in your browser in fifteen minutes.
We will price your real task mix, weigh direct Claude against buying through Microsoft, and set the governance and commitment terms before Microsoft sizes them for you.
One email a month on Copilot Credits, EA renewal levers, Azure commitment, and audit defense. Buyer side only.
