Zapier AI Steps Cost 1, 3, or 5 Tasks: The 2026 Cost Math Before You Build

Zapier charges 1 task per AI step on Standard models, 3 on Advanced, and 5 on Premium, with bring-your-own-model at 1 task, per Zapier’s rates page. Make’s pricing FAQ and operations doc give no AI credit multiplier. At 300 runs a month, the model tier you pick moves the bill from 1,500 to 2,700 tasks — so do the math before you build.

This guide is based on official Zapier and Make documentation and pricing pages. We did not run either tool hands-on.

What a Zapier AI Step Actually Costs: 1, 3, or 5 Tasks

A Zapier AI step costs 1 task per step for Standard models, 3 tasks for Advanced models, and 5 tasks for Premium models, while bringing your own model costs 1 task, according to Zapier’s pricing rates page. AI tool calls are billed separately because the agent picks tools at runtime.

That rate card converts a product decision (which model do I use?) into a line item on your monthly bill. The full rate card from that page:

Standard AI models: 1 task per step, with no tool calls on this tier.
Advanced AI models: 3 tasks per step.
Premium AI models: 5 tasks per step.
Bring your own model: 1 task per step.
Per AI tool call: billed separately, since the agent decides which tools to use at runtime.
Lead Router: 5 tasks per step.
Code step (Code by Zapier, AI-generated or hand-written): 1 task per step, plus 1 task per 30 seconds of extended runtime beyond your plan allowance.
MCP tool call: 2 tasks per call. SDK/CLI actions are listed as “Free in beta.”
Built-in tools (Formatter, Paths, Filters, Delay, Looping, Digests, Zapier Manager, Storage, Tables, Forms): 0 tasks. Sub-Zaps are not in that list: the task-usage article says a Call a Sub-Zap step, a Return From a Sub-Zap step, and each action step inside a sub-Zap cost 1 task per successful run, even though the rates page lists Sub-Zaps as built-in.

The expensive line items are the ones where a model or an agent makes a judgment call at runtime — you cannot predict the tool-call count from the editor, only from the run. Deterministic steps are not all free: Lead Router costs 5 tasks per step, and extended-runtime code costs 1 task per 30 seconds beyond your plan allowance. That asymmetry should drive how you design the Zap, not just how you budget it.

How Zapier Counts Tasks, and What Never Counts

Zapier counts 1 task per successful action step only: triggers never count, polls never count, filters and paths don’t count, and errored or halted steps don’t count, per Zapier’s task-usage help article. Built-in tools like Formatter and Paths cost 0 tasks.

The page’s own worked example: a Zap with a Gmail trigger, a Dropbox action, and a Slack action costs 2 tasks per run, so 10 runs cost 20 tasks. The trigger is invisible to the meter. So are the steps that never executed because a filter blocked them or an earlier step errored.

The page states the principle plainly: “A task is any successful action that runs in Zapier. Only successful actions count toward your task usage.” It also states that polling triggers “never use tasks,” which is the biggest structural difference from Make’s operation model: a poll costs nothing on Zapier, including the 2-minute polling interval the Zapier pricing page lists for Professional.

Two carve-outs matter for budgeting. Testing steps in the editor do not count, so you can iterate without burning tasks. And Zapier Agents usage does not count as tasks at all — Agents run against a separate activity quota, so an agent-heavy build is a different billing conversation than a Zap-heavy one. If you are comparing error and retry behaviour across tools, our breakdown of what retries cost on Zapier, Make, and n8n covers the failure paths that the happy-path docs skip.

The Cost Math Before You Build (Our Calculation)

At Zapier’s published rates, a four-action Zap with one AI step costs 5 tasks per run on a Standard model, 7 on Advanced, and 9 on Premium; at 300 runs a month that’s 1,500, 2,100, or 2,700 tasks — our calculation from Zapier’s rates page, not a vendor claim.

Here is every step, with assumptions stated.

Scenario S1 — deterministic, no AI step. 1 trigger plus 4 action steps.

Calculation, Zapier: trigger = 0 tasks (triggers never count, per the task-usage article). 4 action steps × 1 task = 4 tasks per successful run.
Calculation, Make: 1 trigger operation + 4 module operations = 5 credits per run, assuming exactly one bundle arrives per run and each module costs 1 credit (the Make pricing FAQ says “most actions consume 1 credit”).

Scenario S2 — S1 plus one AI step.

Calculation, Zapier Standard: 4 + 1 = 5 tasks per run.
Calculation, Zapier Advanced: 4 + 3 = 7 tasks per run.
Calculation, Zapier Premium: 4 + 5 = 9 tasks per run.
Make: the pricing FAQ says AI features “may use more” credits but publishes no multiplier, so no number can be stated. That is an open question on the fetched pages.

Volume — 10 new leads per day, about 30 days, so roughly 300 runs per month.

Calculation, S1 on Zapier: 300 × 4 = 1,200 tasks/month.
Calculation, S1 on Make: 300 × 5 = 1,500 credits/month.
Calculation, S2 on Zapier Standard: 300 × 5 = 1,500 tasks/month.
Calculation, S2 on Zapier Advanced: 300 × 7 = 2,100 tasks/month.
Calculation, S2 on Zapier Premium: 300 × 9 = 2,700 tasks/month.

Polling asymmetry.

Calculation, Make: a trigger polling every 15 minutes runs 96 checks per day (24 × 60 ÷ 15) ≈ 2,880 checks per month, if each check consumes 1 operation. The operations doc says a trigger module “runs once to check for new data -> 1 operation,” but it gives no zero-bundle example, so treat this as a scenario to test on a free workspace, not a bill.
Calculation, Zapier: 2-minute polling on Professional = 30 checks per hour × 24 = 720 per day × 30 ≈ 21,600 checks per month at 0 tasks, because polling never counts.

The plan anchors you are measuring against: Zapier Free is $0/month with 100 tasks and two-step Zaps only; Professional starts from $19.99/month; Team starts from $69/month, per the Zapier pricing page. That same page notes that “the higher the task tier, the lower the cost per task,” which matters once your S2 Premium number lands at 2,700 tasks a month — 27 times the Free plan’s entire allowance. Our Zapier vs Make billing differences walkthrough covers how each meter is structured beyond these headline numbers.

The Make Side: Credits, Operations, and Two Open Questions

Make bills in credits: most actions cost 1 credit, the router and error handlers cost none, and AI features via Make’s AI Provider “may use more” without a published multiplier, per Make’s pricing FAQ — so Zapier’s per-model rates are the only auditable numbers here.

What the fetched pages do establish:

Free plan: “If you are on the Free plan, you have 1000 credits per month.”
Credits are the billing unit as of August 27, per the pricing FAQ: “Each action your scenario performs consumes a certain number of credits - most actions consume 1 credit, while some advanced features leveraging Make’s AI Provider may use more.”
Free modules: the router module consumes no credits, and the error handler modules (Rollback, Break, Resume, Commit, Ignore) consume no credits.
Operations definition, from Make’s operations doc: “An operation is a single module run to process data or check for new data.” A module’s operation count equals the number of bundles it processes.
Trigger exception: trigger modules “only run once to check for or retrieve data, regardless of the number of bundles returned.” The doc’s example: Google Sheets Watch New Rows = “1 operation (one per check).”
Worked example: a trigger returning 10 bundles plus three further modules each running 10 times = “Total: 31 operations (1+30) for each scenario run.”
Aggregation lever: replacing 10 separate Gmail sends with one aggregated send cut a run from 11 operations to 3.

Per-plan dollar amounts did not render on the fetched Make pricing page, so no Make prices are stated here.

Two questions stay open. First, the AI credit multiplier: the FAQ says “may use more” and stops there, so a Make AI step cannot be priced from the Make pages fetched for this guide. Second, whether a trigger check returning zero bundles still consumes a credit — the operations doc says one operation per check but never gives a zero-bundle example, so that is a scenario to test, not a fact to plan on.

Line item	Zapier (tasks)	Make (credits)
Standard action step	1 task per successful run	1 credit per operation (most actions)
AI step	1 (Standard) / 3 (Advanced) / 5 (Premium) / 1 (BYO model)	“More” than 1 — multiplier not stated on fetched pages
Trigger	0 tasks	1 operation per check (credit mapping unconfirmed)
Polling for new data	0 tasks	1 operation per check — zero-bundle billing is an open question
Branching / router / filter	0 tasks (Paths, Filters)	0 credits (router module)
Error handling	1 task per successful step in an error handler path	0 credits (Rollback, Break, Resume, Commit, Ignore)
AI tool calls	Billed separately per call; rate not stated on fetched pages	Not stated on fetched pages
MCP tool call	2 tasks per successful call	Not stated on fetched pages

Lock-In Warnings Written Into Zapier’s Own Pages

Zapier’s rates page states “Rates may change over time as we introduce new capabilities or update how tasks are measured,” and the Zapier pricing page says pay-per-task billing is “charged at a higher per-task rate than your base subscription tasks” — two published reasons to price in headroom before you build, both.

The page says the measurement can change, so the 1/3/5 rate card you budget against in 2026 is not guaranteed for the life of your automation. The page adds that changes will be announced before they apply — useful, but an announcement is not a veto.

The overage mechanics carry the same warning. Neither public page gives a number for the pay-per-task rate, so an overrun is a cost you cannot pre-compute. The help article does note that you are notified at 100% of pay-per-task billing, defined as 3x the selected plan’s task limit, after which other tasks are held until the cycle ends. Held runs can be released by replaying them or upgrading your plan.

Make’s version of lock-in is expiry rather than repricing. Per the Make pricing FAQ, unused credits expire at the end of your term, and extra credits expire after one month on monthly plans or one year on annual Pro and Teams plans. Credits you bought and did not spend are not an asset you carry forward.

The scheduled change to watch: starting December 1, 2026, Next Gen Zaps and the Zapier SDK also use tasks, and each completed Next Gen run uses at least 1 task, per the Zapier task-usage article. That is scheduled, not in force at the time of writing. If your build leans on the SDK because it was listed as free in beta, model the post-December cost now. Before committing either way, run a platform exit audit of what you can actually export so a rate change is a migration decision rather than a hostage situation.

FAQ

Builders ask the same four questions before committing: which Zapier steps are free, whether failed runs cost tasks, what changes on December 1, 2026, and whether Agents and MCP calls count — answered only from Zapier’s official pricing and help pages.

Do failed or filtered-out Zap steps cost tasks?

Only successful steps count. Per Zapier’s task-usage article, “Only successful actions count toward your task usage.” Action steps that error or halt do not count, and steps that never run because a filter or path blocked them do not count either. Editor tests are also free of task charges. Successful steps in an error-handler path, and Search actions set to proceed when nothing is found, each cost 1 task.

Does Zapier charge tasks just to check for new data?

No. The Zapier task-usage article says polling triggers “never use tasks.” Make’s operations documentation describes one operation per trigger check; whether a zero-result check is billed is untested.

Which built-in Zapier tools cost 0 tasks?

Per Zapier’s rates page, these built-in tools cost 0 tasks: Formatter, Paths, Filters, Delay, Looping, Digests, Zapier Manager, Storage, Tables, and Forms. Sub-Zaps are the exception, as the task-usage article charges their call and return steps. The two Zapier pages disagree on Sub-Zaps, so budget them as paid until Zapier resolves it.

What changes on December 1, 2026 — and do Agents or MCP calls count?

Starting December 1, 2026, Next Gen Zaps and the Zapier SDK are scheduled to use tasks, with each completed Next Gen run using at least 1 task, per Zapier’s task-usage article. Zapier Agents do not consume tasks — they use a separate activity quota — while MCP tool calls cost 2 tasks each.

The Bottom Line

Pick Zapier if you can pin AI steps to Standard or bring-your-own models at 1 task each (Zapier rates page) and want polls that never use tasks; pick Make if your flows rely on error-handler directives, which consume no credits, since neither fetched page prices the work inside an error route (Make pricing FAQ), while Zapier bills successful steps inside an error-handler path — and either way, budget for rate changes you can’t vote on.

The model tier is the largest lever you control on Zapier: dropping from Premium to Standard on our 300-run scenario cuts 1,200 tasks a month. Make’s advantage is structural rather than published — free error-handler modules and 1,000 free credits — but its AI cost is unauditable from the pages fetched here, so the comparison turns on forecastability, which Make’s public pages do not settle.

How This Guide Was Built

Every vendor figure here comes from official Zapier and Make pages, including the Zapier pricing page and rates page, and Make’s pricing and operations documentation; the authors did not run either tool hands-on, and every arithmetic line is labeled as a calculation rather than a vendor claim.

Where the documentation is silent, this guide says so instead of estimating. That covers Make’s AI credit multiplier, Make’s per-plan dollar amounts (the page did not render them), and whether a zero-bundle Make trigger check consumes a credit. The December 1, 2026 Next Gen and SDK task change is presented as scheduled, not as current billing.

ToolBrain — tool reviews, LLM comparisons, and AI workflow guides

Cross-links automatically generated from NoCode Insider.