Verdict: Zapier changed its pricing page on October 30, 2026 so that AI steps, code steps, MCP and the SDK all draw from one shared task pool, and billing for Next Gen Zaps and the SDK starts at 12 pm ET (17:00 UTC) on December 1, 2026. Independent reporting makes the same point: your plan price stays put while task usage can rise. For any AI step that only classifies, summarises or drafts text, the Standard tier (1x) beats the Premium default (5x). Run the 30-minute audit below this week, before December 1.
TL;DR
- Last verified: 3 November 2026.
- Zapier AI steps now cost tasks by model tier: Standard 1x, Advanced 3x, Premium 5x, Bring Your Own Key 1x.
- New AI steps on paid plans default to Premium, so unchanged habits quietly cost five times as much.
- In the worked example below, an 800-leads-a-month triage Zap falls from 5,600 tasks to 2,400 by switching one dropdown.
- Next Gen Zaps and SDK usage leave their free early-access period on December 1.
How does Zapier AI step task pricing work from December 1?
Zapier's pricing page states that every step in a Zap and every external connector call uses tasks, and that AI model tier, code runtime and connector type each affect how many tasks an action uses. The same page says Zaps, AI steps, code, MCP and SDK share one allocation, with no separate budgets by product. One MCP tool call uses two tasks.
The multiplier comes from the tier you pick, as set out in Zapier's AI by Zapier model tier article (updated November 2, 2026):
| Tier | Task multiplier | Tool support | Notes |
|---|---|---|---|
| Standard | 1x | No | Cheapest; cannot call tools |
| Advanced | 3x | Yes | Auto model is Claude 4.5 Haiku |
| Premium | 5x | Yes | Default for new AI steps on paid plans; Auto model is Claude 4.6 Sonnet |
| Bring Your Own Key | 1x | Yes | Advanced or Premium models only; you pay your provider directly |
The formula is: tasks per run = (1 x model rate) + (tool calls x model rate). A Premium step with two tool calls therefore uses 15 tasks per run, while the same step on Standard with no tools uses 1.
Two dates matter. From December 1, AI by Zapier uses its own price when run through Zapier MCP or the Zapier SDK. Separately, per Zapier's billing article, each app action and fetch call in a Next Gen Zap uses 1 task and each successful run uses at least 1. Next Gen Zaps were free during early access, and that ends for everyone at once. Zapier Tables and Zapier Forms remain free.
Legacy AI steps (created before June 15, 2026) are never updated automatically and cannot use tools; those pairing knowledge sources with a Standard model were paused on June 30, 2026.
Why do agencies feel the Zapier AI step change first?
Agencies tend to build one pattern and clone it across clients. If the template contains a Premium AI step with tools, every client copy inherits the 5x rate and the tool-call multiplication. Ten clients with the same template means ten places where the default is doing work that a cheaper tier could do.
The second pressure is invisibility. The monthly plan price does not move, so nothing looks different on an invoice until the task counter does. The Back Office Beat report frames it the same way: a stronger model means extra tasks for the same Zap, and overages bill at a higher per-task rate. Third-party tracker StackScout reports overage at 1.25x the tier's per-task price on annual billing, and Latenode is cited there for 2.5x on monthly billing. Treat both as third-party figures.
For reference, StackScout lists Professional at $19.99 a month billed annually and Team at $69. At the market rate of 94.93 INR to the dollar on November 3, 2026, that is roughly ₹1,898 and ₹6,550. This is a conversion, not an official Zapier India price.
How do you run the 30-minute pre-December audit?
Work through these five steps in order; budget about six minutes each.
- Find the AI and code steps. Open Zap History, filter runs by your newest AI or Code steps, and note the tasks each step actually consumes per run. Zapier says run history shows tasks consumed per run.
- Apply the formula. For each AI step, calculate (1 x tier rate) + (tool calls x tier rate). Write the result beside the step name.
- Fix the default. If a step only classifies, summarises or drafts text and uses no tools, switch it to Standard. Premium is only worth paying for when the step genuinely needs tools or stronger reasoning.
- Check your billing posture. Confirm which Zaps are Next Gen Zaps or SDK-based, since those start billing on December 1. Decide deliberately whether overage is on. Keep the per-run task limit enabled: the default of 75 tasks (raisable to 500) pauses a step and asks for approval, which guards against a runaway agent.
- Consider Bring Your Own Key. For high-volume tool workflows, BYOK bills at 1x tasks while you pay your model provider directly. That moves the cost rather than removing it, so compare your provider's rates first.
Code steps deserve a glance in step 1: StackScout reports Code by Zapier costs 1 task plus 1 for each 30 seconds beyond the plan's included runtime, so long-running scripts are the ones to check.
Worked example: an 800-leads-a-month triage Zap
This is arithmetic from the published formula, not measured data. The flow is: AI step, then one CRM action, then one Slack action, with no tool calls in the AI step.
| AI step tier | AI step tasks | CRM + Slack | Tasks per run | Tasks per 800 leads |
|---|---|---|---|---|
| Premium (default) | 5 | 2 | 7 | 5,600 |
| Advanced | 3 | 2 | 5 | 4,000 |
| Standard | 1 | 2 | 3 | 2,400 |
| BYOK (Advanced or Premium model) | 1 | 2 | 3 | 2,400 |
Moving from the Premium default to Standard cuts that one Zap's monthly task use by more than half, per the table above. Compare the totals with the task allocation on your own plan, since that is what determines whether you reach overage.
Frequently asked questions
Q: Do all Zapier AI steps cost 5 tasks?
A: No. Cost depends on the tier: 1x for Standard, 3x for Advanced and 5x for Premium, though new AI steps on paid plans default to Premium (Zapier help).
Q: Do triggers and failed actions use tasks?
A: No, only successful actions count. Per Zapier's pricing page and StackScout, failed steps and built-in tools such as Filter, Formatter, Paths, Tables and Forms cost nothing.
Q: What happens when one AI run hits 75 tasks?
A: The step pauses and waits for your approval before continuing. The default limit of 75 can be raised to 500 (Zapier help).
Q: Can I keep paying nothing for AI steps after December 1?
A: No. Through MCP or the SDK, AI by Zapier uses its own rate from December 1, and Next Gen Zaps leave free early access the same day, though Tables and Forms stay free (billing article).
Q: Is there a way to cut AI step costs further?
A: Bring Your Own Key bills at 1x tasks and you pay your own provider directly, but it only works with Advanced or Premium models (Zapier help).
Where can you read more?
If you are weighing Zapier against other builders, see our comparison of Zapier and Lindy for AI agents. For another view of connecting AI to your stack, read AI Connect compared in 2026. If your team writes its own code steps, agentic AI coding tools in 2026 covers the options. Agencies also tidying up billing software may like Zoho Invoice versus its rivals.
Corrections log: none yet. This article was drafted with AI assistance and checked against Zapier's published documentation on 3 November 2026. Read our editorial policy at how we work.
Top comments (0)