Lambda can now checkpoint, suspend for up to a year and resume where it left off, in ordinary Python, TypeScript or Java. That puts a third option between Step Functions and Temporal. How each one models a workflow, what each costs, and which kind of team each suits.
In December 2025 AWS added durable functions to Lambda. You wrap a handler with the Durable Execution SDK, put each side effect in a step, and Lambda checkpoints the result. A wait suspends the function with no compute charge. If the function crashes or resumes after a wait, it replays from the top and skips every step that already completed. An execution can run for up to a year.
If that sounds like Temporal, it is the same idea: durable execution by replay. The difference is that it lives inside Lambda, with nothing to run.
Durable functions: normal Lambda compute and requests, plus $8.00 per million durable operations (starting executions, steps, waits), $0.25 per GB of checkpoint data written and $0.15 per GB-month retained. Retention after completion is configurable from 1 to 90 days, 14 by default.
Step Functions Standard: $0.025 per 1,000 state transitions, which is $25 per million. Express workflows bill per request and duration instead, for high-volume work under five minutes.
Temporal: Temporal Cloud bills by actions and storage; self-hosting is free software plus a database, a cluster and someone on call for it.
A ten-step workflow costs roughly $0.25 per thousand runs in Step Functions transitions, and roughly $0.09 per thousand in durable operations, before compute. Step Functions often calls services directly, with no Lambda in between. Durable functions always pay Lambda compute for the steps. Which one is cheaper depends on how much of the work is your code.
Book a free 30-minute call. You'll leave knowing what we'd do first and roughly what it costs, whether or not you hire us.
Durable functions and Temporal share a constraint that Step Functions does not have: workflow code must be deterministic. On replay, the code runs again from the beginning, and it has to make the same decisions in the same order. That means:
Anything with a side effect or a non-repeatable result (an API call, a timestamp, a random number, a UUID) goes inside a step.
Changing workflow code while executions are in flight needs care. A deploy that reorders steps can break replay of executions started on the old code. Publish versions and let old executions finish on the version they started with.
Tests have to cover replay, not only the happy path. The SDKs ship local testing tools for this.
Durable functions: a team already on Lambda whose workflow is mostly its own logic. Order processing, AI agent loops with human approval, multi-step jobs that used to need a state table and a scheduler. The least new infrastructure of the three.
Step Functions: workflows that are mostly calls to AWS services, where people other than the authors need to read the flow, or where you want zero runtime maintenance. Also the cheaper choice when most steps are direct service integrations.
Temporal: workflows that span clouds or run outside Lambda, very long-lived entities, or teams that want the same engine in every environment. You pay for it in operations or in Temporal Cloud fees.
Durable functions launched in US East (Ohio) for Python 3.13 and 3.14 and Node.js 22 and 24, and expanded through 2026, including 16 more Regions in April. Java support (17 and later) followed the February developer preview, and the .NET SDK reached general availability in July. Check the Region list before you design around it.
Once a month: what changed in tech that month and what it means for you, from cloud and AI to security and the tools you build with. Every item links to its source. No sales pitch, unsubscribe in one click.
Book a free 30-minute call about it. Bring your setup and your bill; you'll leave knowing what we'd do first and roughly what it costs, whether or not you hire us.
Or keep reading
AWS Lambda Is Five Things Now: Event Functions, Web Functions, Durable Functions, MicroVMs and Managed Instances
In twelve months Lambda grew from one compute model to five. Event Functions (the classic model, renamed), Web Functions for HTTP servers billed on active CPU, Durable Functions for long workflows, MicroVMs for isolated sandboxes, and Managed Instances on EC2. What each is for and how each bills.
Read
Lambda Cold Starts With Postgres, RDS Proxy and DynamoDB, Measured: VPC or Not, 200 at Once
Node.js 22 functions on arm64 took 570–650 ms end to end on a cold start, whether they talked to Postgres directly, through RDS Proxy, or to DynamoDB, and a function in a VPC was no slower than one outside it. Bursts of 200 simultaneous cold starts all succeeded. Warm, RDS Proxy added about 3 ms per query.
Read
As of October 25, 2026. Provided as is, for general information, and accurate to the best of our knowledge on the date shown. Not legal, compliance, tax, financial or other professional advice. Prices, limits and vendor services change; check current vendor documentation before relying on anything here. Terms of use.