Microsoft Copilot Cost Calculator

Estimate Microsoft Copilot Cowork costs in Copilot Credits by user persona and task mix. Compare pay-as-you-go vs pre-purchase (P3) pricing. Free tool.

Advertisement

Microsoft Copilot cost calculator for consumption-based Copilot Credits

This calculator models what Microsoft Copilot Cowork costs you per month once people actually start running tasks with it. It is not a per-seat licence estimator. It takes the number of users in each persona, how many tasks they run per workday, how heavy those tasks are, and turns that into a monthly Copilot Credit burn and a dollar figure — then shows what committing to a pre-purchase plan does to the annual number.

Everything runs in your browser. Nothing you type is sent anywhere, there is no sign-in, and the “Copy estimate” button writes a plain-text summary to your clipboard rather than emailing it. That matters if the headcount and task-volume numbers you are modelling are not public yet.

The one price the tool ships, and what it does not ship

The calculator hard-codes exactly one price: $0.01 per Copilot Credit, the pay-as-you-go rate. Every dollar figure on the calculator tab is that rate multiplied by a credit count you control. The tool ships no per-user licence prices at all — no Microsoft 365 Copilot seat price, no E3/E5 price, no Business Premium price. If you are looking for a per-seat licence total, this is the wrong page; the credit model is a different axis of spend entirely.

Two honest caveats about that $0.01. First, it is the figure the tool ships, and Microsoft list pricing changes — treat it as the assumption baked into the model, not as a quote. Second, it is the pay-as-you-go rate specifically; the pre-purchase discount slider is how you model anything cheaper.

The credit costs per task are not fixed prices — they are inputs you can edit. The tool ships defaults of 200 credits for a light task, 550 for medium, and 900 for heavy, and its tooltips describe Microsoft's published grading bands as roughly 100–300 light, 400–700 medium, and 700+ heavy. The defaults sit inside those bands. If you have observed consumption from a pilot, replace them — the whole estimate scales linearly off those three numbers.

The prerequisite people miss

Copilot Credits are consumption on top of licensing, not instead of it. Cowork requires a Microsoft 365 Copilot licence per user, and that subscription is billed separately from every credit this calculator counts. So the real budget line is two things added together:

  • Per-user Microsoft 365 Copilot subscription — a fixed monthly seat cost, not modelled here.
  • Copilot Credits consumed by Cowork tasks — variable, and what this calculator estimates.

A plan that budgets only for seats will be short by the entire output of this tool. A plan that budgets only for credits will be short by the seat cost. Both numbers are real.

It is also worth understanding what drives a credit charge, because it explains why the light/medium/heavy grading exists at all. A task's credit cost comes from the model it uses, the context it retrieves, the tool calls it makes, and how long it runs — not from a flat per-message rate. Two prompts that look identical to a user can differ several-fold in credits if one of them pulls in a large document set and calls three connectors.

The persona model, and how to fill it in

Rather than asking for one blended usage number, the calculator splits your population into four editable personas, because task mix varies far more between roles than volume does within a role. The shipped defaults are a 200-user organisation:

PersonaUsersLight/dayMedium/dayHeavy/day
Corporate Knowledge Workers100410
Customer-Facing Workers30620
Technical Workers20321
Managers & Senior Leaders50310

Every cell is editable, including the user counts. Task-per-day fields step in halves, so “one heavy task every other day” is 0.5. The two remaining assumptions are working days per month (default 21) and the pre-purchase discount percentage (default 15).

The arithmetic for each persona is deliberately transparent:

  • Credits per user per day = (light tasks × light credits) + (medium × medium credits) + (heavy × heavy credits)
  • Monthly credits = users × working days × credits per user per day
  • Monthly cost = monthly credits × $0.01

A worked example using the shipped defaults

Run the numbers above with 200 credits light, 550 medium, 900 heavy, and 21 working days:

PersonaCredits/user/dayMonthly creditsMonthly cost
Corporate (100 users)4×200 + 1×550 = 1,3502,835,000$28,350
Customer-facing (30)6×200 + 2×550 = 2,3001,449,000$14,490
Technical (20)3×200 + 2×550 + 900 = 2,6001,092,000$10,920
Leaders (50)3×200 + 1×550 = 1,1501,207,500$12,075
Total (200 users)6,583,500$65,835

That is $790,020 a year pay-as-you-go, and 32,917.5 credits per user per month. Apply the default 15% pre-purchase discount and the annual figure becomes $671,517, a saving of $118,503. The discount is applied to the annual total only; the monthly card on the pay-as-you-go side is unchanged.

Notice how much of that total comes from 20 technical users. They are 10% of headcount and 16.6% of the credits, because one heavy task a day costs more credits than four light ones. This is the single most useful thing the persona split reveals: your credit bill is driven by a small population running expensive tasks, not by broad adoption of cheap ones.

The pre-purchase plan slider — what it does and does not model

The pre-purchase plan (referred to in the tool as P3) is a one-year commitment with tiered discounts; the tool's own note describes the tiers as reaching up to roughly 20%, and overage beyond your commitment spills back to pay-as-you-go rates. The slider is a flat percentage off the annual pay-as-you-go total. It does not model tier thresholds, it does not model the overage spill, and it does not know what commitment level your discount corresponds to.

Practically, that means the P3 figure is a floor-of-optimism number: it assumes you commit accurately. If your real consumption overshoots the commitment, the excess is billed at full pay-as-you-go rate and your effective discount is lower than the slider says. Model that by running the tool twice — once at your committed volume with the discount, once at your expected volume with 0% — and treating the truth as somewhere between.

The model routing optimizer tab

The second tab answers a different question: given the annual credit volume from tab one, how much does it matter which model tier the work runs on? It defines three tiers with editable relative multipliers, where the premium tier is the 1.0 reference point:

TierExample named in the toolDefault multiplier
LowCowork 1 / lightweight0.30×
MidSonnet 4.60.60×
HighOpus 4.81.00×

The baseline is everything running on the premium tier. The optimised figure weights your annual credits by a routing mix you set (defaulting to 30% low / 60% mid / 10% high), and the mix is normalised automatically if the three percentages do not sum to 100.

With the default persona set and the default mix: annual credits are 79,002,000, the baseline at 1.0× is $790,020, and the blended factor is (0.30×0.30) + (0.60×0.60) + (0.10×1.00) = 0.55. The optimised annual cost is $434,511 — a saving of $355,509, or 45% lower.

Be careful with that number. The tool is explicit that Microsoft does not publish a per-model credit rate, so the 0.3 / 0.6 / 1.0 multipliers are planning estimates mirroring the relative price hierarchy of the available models, not measured consumption. The savings percentage is therefore a sensitivity analysis, not a forecast. What survives the uncertainty is the direction: if premium reasoning models cost meaningfully more credits per task than lightweight ones, routing routine work away from them is the largest single lever on Cowork spend, and the exact multipliers only change how large.

Also note the two tabs are not additive. The optimizer's baseline is deliberately identical to the calculator's pay-as-you-go annual total whenever the high-tier multiplier is left at 1.0 — it is the same money, re-cut by model tier. Do not add $790,020 and $434,511 together.

Failure modes and sanity checks

  • Task counts are per user per workday, not per month. Entering 80 in the light column because “that's about four a day” inflates the estimate roughly twentyfold.
  • Working days, not calendar days. The default of 21 is a working month. Putting 30 in that field silently adds about 43% to your bill.
  • Blank or non-numeric entries become zero. A persona whose user count is cleared contributes nothing rather than erroring, so a total that looks suspiciously low is usually an emptied field.
  • Credits per task are the biggest lever and the weakest input. Every result scales linearly with them, and they are the numbers you are least likely to know before a pilot. Run the model at the low and high ends of the published bands to see the range you are actually committing to.
  • The heavy column deserves scrutiny. At the shipped defaults one heavy task costs 4.5 light tasks. If any persona's heavy figure is a guess, it is the guess most worth replacing with measurement.
  • Nothing persists. Reloading the page resets every field to the defaults. Use “Copy estimate” before you close the tab — it captures the assumptions alongside the totals, which is what makes a saved estimate reviewable later.

Using the output in a budget conversation

The copied estimate deliberately leads with assumptions before totals, because the totals are only defensible if the assumptions are visible. When you take this to a finance review, present three things: the credit price you assumed, the per-task credit grading you assumed, and the task volumes per persona. The first is a published rate you can point at; the second and third are your organisation's estimates and are where every argument will actually happen.

The most productive way to use the tool is not to produce one number but to bracket a range — conservative task volumes with low-band credits, aggressive volumes with high-band credits — and then decide whether the gap between those two figures is small enough to commit against. For most organisations it is not, which is the argument for running a metered pilot on pay-as-you-go before signing a one-year pre-purchase commitment.

Frequently Asked Questions

How is Microsoft Copilot Cowork priced?+

Copilot Cowork uses usage-based billing measured in Copilot Credits. Pay-as-you-go costs $0.01 per credit, with no commitment. Credits are consumed per task based on model use, context retrieval, tool calls, and runtime — so a task's cost depends on its complexity, not a flat per-message rate. Organizations can also pre-purchase credits (the P3 plan) for a discount.

What are Copilot Credits?+

Copilot Credits are Microsoft's common currency for usage-based AI services such as Cowork and Work IQ. Each credit costs $0.01 on pay-as-you-go. Instead of paying per message, you pay for the actual work a task does — heavier tasks that use more capable models, retrieve more context, or call more tools consume more credits.

How many credits does a Cowork task use?+

Microsoft grades tasks into three tiers: light tasks (roughly 100–300 credits) draw on a small number of sources with limited reasoning; medium tasks (roughly 400–700 credits) use multiple sources and structured reasoning; and heavy tasks (700+ credits) aggregate broadly and apply deep reasoning. This calculator lets you set the exact credit value for each tier to match your own observed usage.

What is the Copilot pre-purchase plan (P3)?+

The pre-purchase plan (P3) is a one-year, pay-up-front commitment to a volume of Copilot Credits in exchange for a tiered discount of up to about 20% versus pay-as-you-go. Prepaid credits are drained first, and any usage beyond your commitment automatically continues at the pay-as-you-go rate. The calculator's comparison shows your projected annual savings at whatever discount you enter.

Do I still need a Microsoft 365 Copilot license to use Cowork?+

Yes. Copilot Cowork requires a Microsoft 365 Copilot user subscription license as the entry point. That per-user subscription is billed separately from the Copilot Credits consumed by Cowork tasks, so your total cost is the license fee plus usage-based credit spend.

How can I reduce Copilot Cowork costs?+

The biggest lever is model choice: premium reasoning models (such as Opus 4.8) cost the most credits per task, while mid-tier (Sonnet 4.6) and lightweight models cost a fraction. Reserving premium models for genuinely complex work and routing routine tasks to cheaper tiers can cut spend substantially — the Model Routing Optimizer tab quantifies this. Microsoft also provides spending limits, usage alerts, and per-user controls in the Microsoft 365 admin center to prevent overspend.

Is this calculator an official Microsoft quote?+

No. This is an independent planning tool. Pay-as-you-go pricing ($0.01/credit) and the light/medium/heavy task ranges are based on Microsoft's published documentation, but actual credit consumption varies by workload, model, and configuration. Microsoft does not publish a fixed credit cost per model, so the routing multipliers are estimates. Use the tool to model scenarios and budgets, then confirm with Microsoft's admin-center reporting or your licensing partner.

This tool is provided for informational and educational purposes only. All processing happens in your browser — no data is sent to or stored on our servers. While we strive for accuracy, we make no warranties about the completeness or reliability of results.