Serverless platforms run your code on demand without a server you manage, and bill by requests, CPU time or memory.
This guide ranks fourteen of them by vendor independence first, then by setup effort, what a busy month adds to the bill, how easily code moves elsewhere, and whether the platform suits small web apps, edge scripts or heavy compute.
Vendors can pay for more visibility on this website, for example a higher place in a ranking or more room on a page. Commercial relationships can also affect which products we feature and how prominently. Every vendor link is marked the same, whatever the arrangement behind it. How that works.
In short
What serverless platforms do
A serverless platform runs functions or small apps on demand, scales them up and back to zero, and charges for requests, CPU time and memory used rather than idle servers.
Every serverless platform was judged on five points in a fixed order: setup effort, cost in practice, how data is exported, how easily you can leave, and the buyer it was designed for. A separate rule reserves the first three places for challengers and moves the household names to place 4 and below. Why those five, and why there is no score out of ten, is on the how we work page.
Per-execution and GB-second billing on usage plans; capacity-based Premium and Dedicated plans; monthly grant on Consumption; 30-day trial credit
—
Teams building in C#, Java or PowerShell that already use Azure
Buyers who want rates visible before opening an account
The country column shows the vendor's headquarters or contracting entity. Data hosting is a separate matter, and an entry flags it when the two diverge.
App hosting with a free Hobby plan and flat Pro plan
Ranked #1 of 14 in Best Serverless Platforms in 2026.
Free tierPublished pricingElsewhere
The Hobby plan is the draw: 100 apps, 500 deployments, 1,000,000 requests, 150 GB of bandwidth and 50 custom domains at $0, though logs are kept for one day.
Pro costs a flat monthly sum with one seat, daily backups, two-week logs, SSH, SFTP and cron jobs, then charges extra per seat, app and dedicated IP. Templates cover Next.js, Django, Laravel and Astro. The cold-start claim of 107 times faster than Docker comes from the vendor.
What stands out
Hobby plan
Own-infrastructure option
Framework templates
Where it costs you
Active CPU time is capped at 6 hours a month on Hobby, which a busy API could exhaust.
The pages read name no country and no licence for Edge, so procurement checks need extra work.
Right for
Solo developers and small teams deploying framework apps who want fixed monthly limits
Wrong for
Buyers who need a stated legal entity and country upfront
Not stated by the vendorFree Hobby plan; Pro plan billed monthly per seat with listed overage rates; enterprise plan on contact
Small scripts deployed in 100 ms with SQLite, cron and email triggers
Ranked #2 of 14 in Best Serverless Platforms in 2026.
Free tierPublished pricingElsewhere
Limits count runs per day rather than invocations per month: 100,000 on the free plan, 1,000,000 on Pro and 5 million on Business. Pro lifts wall-clock time per run from one minute to ten and tightens cron intervals from 15 minutes to one.
It also adds custom domains and 2,000 private vals. A yearly Pro option shows two months free, and Enterprise brings SOC 2 and log export on custom terms.
What stands out
Cron and email triggers
Built-in SQLite
Daily run quotas
Where it costs you
Runs are capped at one minute on the free plan and ten minutes on Pro, which rules out long jobs.
The pages read show no legal entity, address or country.
Right for
Developers wiring up small cron jobs, webhooks and email handlers without managing infrastructure
Wrong for
Long-running batch work or teams needing contract details upfront
Not stated by the vendorFree plan with daily run limits; Pro and Business plans billed monthly; yearly Pro option; enterprise custom
Usage-priced edge scripts on Deno with scale to zero
Ranked #3 of 14 in Best Serverless Platforms in 2026.
Published pricingElsewhere
Requests cost $0.20 per million and CPU time $0.02 per 1,000 seconds, with a $1 monthly minimum and no minimum capacity. Scripts are TypeScript with npm module access, powered by Deno, and the page claims 24 ms global latency, deploys in under 100 ms and scale to zero.
Bandwidth is the open variable: it is priced separately and no figure appears. The trial lasts 14 days and asks for no credit card.
What stands out
Deno runtime
119 edge locations
Scale to zero
Where it costs you
Bandwidth prices are not shown on the product page, so a full monthly estimate needs another source.
A monthly minimum applies, so light use still pays once the 14-day trial ends.
Right for
Teams adding TypeScript logic at the edge who want billing by requests and CPU
Wrong for
Hobby projects that want an ongoing no-cost allowance
Not stated by the vendorPublished per-request and CPU-time rates with a monthly minimum; bandwidth billed separately; 14-day trial
Apache-2.0 serverless framework running on Kubernetes, Docker or on-premises
Ranked #4 of 14 in Best Serverless Platforms in 2026.
Open sourceSelf-hostablePricing on requestElsewhere
Nuclio is described as both an open source and a managed serverless platform, with the code under Apache-2.0. Stated capabilities include Kubernetes and GPU support, a management dashboard, multi-tenancy and basic monitoring and logging.
The roughly 400,000 events a second quoted is a vendor claim, and explicit limits are absent. Hosting spans Docker, Kubernetes, on-premises, edge and multiple clouds. The footer names iguazio, and no country is given on the pages read.
What stands out
Apache-2.0 licence
Kubernetes and GPU
Multi-tenancy
Where it costs you
The homepage lists no plans, prices or limits, so the cost of the managed option cannot be judged from it.
Running it yourself means operating Kubernetes or Docker, work the hosted entries take over.
Right for
Platform teams that must run functions on their own Kubernetes or on-premises clusters
Wrong for
Buyers wanting a price list and a hosted account in minutes
Not stated by the vendorOpen source code under Apache-2.0; managed option mentioned; no plans or prices on the homepage
Deno hosting with GitHub deploys and a preview domain per pull request
Ranked #5 of 14 in Best Serverless Platforms in 2026.
Free tierPublished pricingNorth America
Plans climb in steps: Free allows 1M requests, 20 GiB egress, 10 active CPU hours and 10 apps; Pro adds 5M requests, 50 CPU hours and 100 custom domains for $20 a month; Builder at $200 a month covers 25M requests and 10,000 apps.
Overage runs $2 per million requests. Enterprise brings a 99.95% SLA. The vendor cites over 10 billion requests a month, and a separate product, deployd, covers customer-run hosting.
What stands out
Preview per pull request
Cron in code
OpenTelemetry tracing
Where it costs you
Free-plan support is community only with no SLA, which matters for anything customer-facing.
A company banner announces joining Cloudflare without terms, so ownership is in flux.
Right for
JavaScript and TypeScript teams that review every change on its own preview URL
Wrong for
Projects needing an uptime guarantee without an enterprise contract
United StatesFree plan; Pro and Builder plans billed monthly with overage rates; enterprise custom with SLA
Container hosting billed per second with scale to zero
Ranked #6 of 14 in Best Serverless Platforms in 2026.
Pricing on requestElsewhere
Pro costs $29 a month plus compute, with $10 included, 10 users, 100 services and email and chat support. Scale costs $299 with $100 included, 1,000 services, AWS Regions and a 99.9% uptime SLA.
Enterprise starts at $1000 a month with SOC 2, ISO 27001, SSO, RBAC and a 99.99% SLA. Cold starts under 200 ms are claimed. A 5-hour serverless Postgres instance and startup credits are offered, but neither is a trial.
What stands out
Per-second billing
Bare-metal VMs
GPU concurrency
Where it costs you
Pro carries a monthly fee before any compute is used, so tiny projects pay a floor.
Joining Mistral AI is announced without terms, and no HQ country is given.
Right for
Teams deploying containers that want per-second billing and scale to zero
Wrong for
Experiments needing a platform plan with no monthly fee
Not stated by the vendorMonthly plan fee plus per-second compute billed at month end; included compute credit; enterprise plan custom
Compute billed per second, with GPUs and endpoints that scale to zero
Ranked #7 of 14 in Best Serverless Platforms in 2026.
Pricing on requestElsewhere
Starter has a $0 base fee, $30 a month in credits, 3 seats, 100 containers and 10 GPU concurrency. Team costs $250 plus compute, with $100 credits, unlimited seats, 5,000 containers and 50 GPU concurrency; Enterprise is custom.
Compute is metered per second: CPU at $0.0000131 per physical core, memory at $0.00000222 per GiB and an H100 at $0.001097. The page describes Spindle as an open-source API but does not call the platform itself open source.
What stands out
Per-second GPU billing
Sandboxes
HTTPS endpoints
Where it costs you
Team costs $250 a month before compute, a steep step up from the Starter plan.
Concurrency is capped per plan, for example 10 GPUs on Starter, so scaling out needs an upgrade.
Right for
GPU and batch workloads that must scale out fast and back to zero
Wrong for
Simple webhooks where a small fixed monthly price matters
Not stated by the vendorStarter plan with monthly credit; compute billed per second; Team plan with unlimited seats; enterprise custom
Deno functions that sit beside a managed database and webhooks
Ranked #8 of 14 in Best Serverless Platforms in 2026.
Free tierPublished pricingAsia-Pacific
Edge Functions are one line in a broader plan. Free covers 500,000 invocations, 50,000 MAUs and a 500 MB database across at most 2 active projects. Pro, from $25 a month, includes 2 million invocations, then $2 per million, 100,000 MAUs and a $10 compute credit.
Team starts at $599 with SOC 2, ISO 27001 and SLAs. Functions scale automatically, can be pinned near the database and deploy through the Supabase CLI with CI/CD.
What stands out
Database webhooks
npm modules
Pinned near database
Where it costs you
Free projects pause after one week of inactivity, which suits prototypes better than steady services.
Compute is charged apart from the plan, so the headline monthly price understates a busy project.
Right for
Apps already built on Supabase that need server code triggered by database changes
Wrong for
Teams wanting functions as a standalone product with simple pricing
SingaporeFree plan with an invocation allowance; Pro and Team tiers billed monthly; overage per million invocations; enterprise custom
Functions bundled into Netlify's credit-based web hosting plans
Ranked #9 of 14 in Best Serverless Platforms in 2026.
Free tierPublished pricingNorth America
Plans are Free with 300 credits, Personal at $9 a month with 1,000, and Pro at $20 with 3,000 credits and unlimited members. Compute costs 10 credits per GB-hour and web requests 2 credits per 10,000.
Background functions return a 202 response at once and run up to 15 minutes, memory goes to 4 GB and 2 vCPUs, and a region can be set per function. Functions come with every plan and deploy with the app.
What stands out
Credit metering
Background functions
Scheduled functions
Where it costs you
Invocation counts and runtime hours are not stated on the pricing page, so budgets rest on credit arithmetic.
The Free plan holds 300 credits in total, a small pool for a site with real traffic.
Right for
Sites already on Netlify that need a few background or scheduled jobs
Wrong for
Buyers who want invocation-based pricing they can forecast directly
United StatesFree plan; Personal and Pro plans billed monthly in credits; function usage metered in credits; enterprise custom
Functions with stated duration, memory and concurrency limits per plan
Ranked #10 of 14 in Best Serverless Platforms in 2026.
Free tierPublished pricingElsewhere
Hobby covers 1M invocations, 4 hours of Active CPU and 360 GB-hours of memory at $0. Pro costs $20 a month with a $20 credit, then invocations from $0.60 per million, Active CPU from $0.128 per hour and memory from $0.0106 per GB-hour.
Bundles can reach 250 MB (500 MB for Python). The default region is iad1, and Pro and Enterprise teams can set several. Temporary deployments expire unless claimed to a Hobby account.
What stands out
Hard-capped Hobby
Active CPU billing
Regional placement
Where it costs you
Hobby cannot buy extra usage, so a traffic spike ends in a cap rather than a bill.
Pro has no cap and three separate meters, so the monthly bill is harder to predict.
Right for
Web teams that want clear per-function limits and regional control
Wrong for
Small projects outgrowing Hobby that need predictable costs
Not stated by the vendorFree Hobby plan with a hard usage cap; Pro plan with included credit then pay-as-you-go rates; enterprise via sales
App deployment on servers from AWS, Hetzner or Linode via an agent
Ranked #11 of 14 in Best Serverless Platforms in 2026.
Free tierPublished pricingElsewhere
Free gives 1 self-managed server, 48-hour logs and a 2C4G build spec. Dev costs $5 a month after a 14-day trial, with 3 servers and 5 domains. Pro costs $19 with 10 servers, 30-day logs, 100 API keys and log forwarding.
Team costs $79 with 3 seats, then $24 per extra seat, unlimited servers and 90-day logs. The Zeabur Agent reads the code and detects language and framework. Nothing on the pricing page covers functions.
What stands out
Server-based plans
Deployment agent
14-day trial
Where it costs you
Plans count servers, not invocations, so per-request serverless use is not metered.
The free plan includes 0 domains and 0 API keys, so a public app needs the Dev plan.
Right for
Developers who want an agent to detect and deploy a whole app
Wrong for
Buyers shopping for per-invocation function pricing and limits
Not stated by the vendorFree plan; Dev, Pro and Team plans billed monthly; 14-day trial on Dev and Pro; enterprise by contract
Workers with a daily-capped free plan and a minimum-based paid plan
Ranked #12 of 14 in Best Serverless Platforms in 2026.
Free tierPublished pricingNorth America
Free allows 100,000 requests a day, resetting at 00:00 UTC, with 50 subrequests and 5 cron triggers.
Paid starts at $5 a month per account, includes 10 million requests and 30 million CPU ms, then charges $0.30 per extra million requests and $0.02 per extra million CPU ms; subrequests rise to 10,000 and cron triggers to 250. Invocations run up to 5 minutes, default 30 seconds. Enterprise follows the customer contract.
What stands out
Daily request cap
CPU-time billing
Cron triggers
Where it costs you
The Free plan allows 10 ms of CPU per invocation, which restricts anything heavier than light routing logic.
Memory is fixed at 128 MB per isolate, so memory-hungry workloads do not fit.
Right for
Light request handling and routing where per-request cost matters
Wrong for
Memory-heavy or long compute jobs beyond the isolate limits
United StatesFree plan with a daily request limit; paid plan with a minimum charge plus request and CPU-time rates; enterprise by contract
Event functions wired to more than 220 native AWS service integrations
Ranked #13 of 14 in Best Serverless Platforms in 2026.
Free tierPublished pricingElsewhere
Event functions cost $0.20 per million requests and $0.0000166667 per GB-second on x86, or $0.0000133334 on Arm; Provisioned Concurrency is $0.0000041667 per GB-second.
Web Functions cost $0.30 per million requests plus $0.0000276944 per vCPU-second, and Managed Instances reach 32 GB and 16 vCPUs. Asynchronous invocations now allow a 90-minute timeout, up from 15 minutes. Examples use US East (N. Virginia) rates.
What stands out
220+ AWS integrations
Provisioned Concurrency
90-minute async timeout
Where it costs you
Several billing models (event functions, Web Functions, Provisioned Concurrency) make cost estimates a multi-step job.
The monthly free tier does not apply to functions with Provisioned Concurrency enabled.
Right for
Teams already building on AWS that need functions triggered by its services
Wrong for
Buyers who want one flat price per function
Not stated by the vendorPublished per-request and per-GB-second rates with a monthly free tier; other models for web functions and managed instances
Functions across Consumption, Flex Consumption, Premium and Dedicated plans
Ranked #14 of 14 in Best Serverless Platforms in 2026.
Pricing on requestElsewhere
Four plans bill differently: Consumption by executions and GB-seconds, with a grant of 400,000 GB-seconds and 1 million executions; Flex Consumption with 100,000 GB-seconds and 250,000 executions; Premium by vCPU-seconds and GB-seconds with no grant; Dedicated per capacity unit.
Flex Consumption and App Service plans carry a 99.95% availability SLA. Hosting also includes Azure Container Apps. A $200 credit lasts 30 days.
What stands out
Four hosting plans
99.95% SLA tiers
Linux Consumption retiring
Where it costs you
Prices are masked on the pricing page read, so costs cannot be compared with other entries.
Linux Consumption retires in September 2028, which forces a plan move for existing users.
Right for
Teams building in C#, Java or PowerShell that already use Azure
Wrong for
Buyers who want rates visible before opening an account
Not stated by the vendorPer-execution and GB-second billing on usage plans; capacity-based Premium and Dedicated plans; monthly grant on Consumption; 30-day trial credit
A serverless platform runs functions or small apps on demand, scales them up and back to zero, and charges for requests, CPU time and memory used rather than idle servers. The differences that matter are rarely in the feature list, so this is the order we would work through them.
01
Check who publishes a price
4 vendors on this page sell by quote, while 10 publish a pricing model you can read today: Wasmer Edge, Val Town, Bunny Edge Scripting, Deno Deploy, Supabase Edge Functions, Netlify Functions, Vercel Functions, Zeabur, Cloudflare Workers, AWS Lambda.
02
Look at where the vendor sits
These 14 vendors are established in 3 countries across 3 regions (Elsewhere 10, North America 3, Asia-Pacific 1). That decides whose disclosure law applies to what the vendor holds, wherever the servers are.
03
Check the open source options
Open source here means Nuclio. The price of self-hosting is the time your team spends maintaining it.
Serverless platforms bill in different units: requests at Bunny Edge Scripting, runs at Val Town, credits at Netlify Functions
A serverless price only means something once you convert it into the unit your workload produces. Bunny Edge Scripting charges per million requests and per 1,000 CPU seconds, Val Town counts runs per day, Netlify Functions meters credits per GB-hour, and Vercel Functions bills invocations, Active CPU and memory as three separate meters. Cloudflare Workers adds CPU milliseconds to requests.
Take one realistic day of traffic, work out requests, CPU time and memory for it, then price that day on each shortlisted platform. Repeat the exercise at ten times the traffic and keep both figures beside each other, because several plans change shape sharply when volume grows and the cheapest option at small scale is often not the cheapest later.
Convert one typical day of traffic into each platform's billing unit
Price the same workload at ten times the volume
Separate request, CPU-time and memory meters before comparing totals
Test the free plan limits of Cloudflare Workers, Supabase Edge Functions and Vercel Functions before production
Entry plans are built to be outgrown, and each one stops in a different way. Cloudflare Workers Free gives 10 ms of CPU per invocation and 100,000 requests a day. Supabase pauses free projects after a week of inactivity. Vercel Hobby is capped and cannot buy extra usage. Val Town limits free runs to one minute, and Wasmer Edge Hobby holds 6 active CPU hours a month.
A prototype can live on any of these, but a customer-facing service needs to know whether the limit ends in an error, a pause or a bill. Read what happens at the limit, not only the number, and test one function against it before launch day arrives, so that the first time you meet the ceiling is not in front of customers.
Find out whether the limit produces an error, a pause or an overage bill
Check CPU time per invocation against your slowest function
Confirm what support the entry plan actually includes
Hyperscaler functions such as AWS Lambda and Azure Functions need a cost model before you commit
AWS Lambda and Azure Functions are the most connected products here, with more than 220 AWS service integrations at one and four hosting plans at the other, but that range makes pricing a multi-step job. Lambda bills event functions, Web Functions, Provisioned Concurrency and Managed Instances differently, and its monthly free tier does not apply to functions with Provisioned Concurrency.
Azure bills Consumption, Flex Consumption, Premium and Dedicated plans on different units, and the pricing page we read masks the rates. If your workload already sits in one of those clouds, the integrations may justify the effort. If not, a smaller vendor with one price page is easier to audit, and a spreadsheet with each billing model as a row will show quickly whether the extra range is worth the extra effort.
List which billing model each function would fall under
Confirm rates are visible before relying on an estimate
Check plan retirements such as Linux Consumption in September 2028
Runtime and ownership changes: Deno Deploy, Koyeb and Nuclio show what to check before moving code
Code portability depends on the runtime, and ownership can shift under a vendor. Bunny Edge Scripting, Supabase Edge Functions and Deno Deploy all run on Deno, so TypeScript with npm modules can move between them with limited rewriting. Nuclio is Apache-2.0 and can run on Docker, Kubernetes or on-premises, which is the cleanest exit if hosted prices change.
Wasmer says it can also run on your own infrastructure. Ownership matters too: Deno's company page says it is joining Cloudflare, and Koyeb's about page says it is joining Mistral AI, both without terms. Check each announcement before signing a long commitment, and ask the vendor in writing what changes for existing customers if a deal closes.
Confirm the runtime your code targets is shared by another host
Read company-page announcements about acquisitions before committing
Keep a deploy path that does not rely on one platform's triggers
What goes wrong most often when buying serverless platforms
Reading a monthly free allowance as a production budget, then meeting a daily cap on Cloudflare Workers or a run limit on Val Town.
Choosing Zeabur for per-request serverless billing when its plans count servers and its pricing page never mentions functions.
Comparing Azure Functions prices without noticing that the rates were masked, which leaves the cheapest-looking option unproven.
Assuming Hobby plans at Vercel Functions or Wasmer Edge can absorb a traffic spike, when both stop at stated limits.
07
Frequently asked questions
8 answers
What are the best serverless platforms in 2026?
Wasmer Edge leads our ranking of 14. Wasmer Edge states its limits in plain numbers: the Hobby plan allows 100 apps, 1,000,000 requests and 150 GB of bandwidth a month, and Pro adds seats, backups, SSH access and cron jobs.
The page adds that Wasmer can be run on infrastructure you own. The pages read name no country, no licence for Edge itself and no length for the Pro trial, so company checks stay with the buyer.
Which serverless platforms publish their pricing?
10 of the 14, with the pricing model each one publishes:
Wasmer Edge: Free Hobby plan; Pro plan billed monthly per seat with listed overage rates; enterprise plan on contact.
Val Town: Free plan with daily run limits; Pro and Business plans billed monthly; yearly Pro option; enterprise custom.
Bunny Edge Scripting: Published per-request and CPU-time rates with a monthly minimum; bandwidth billed separately; 14-day trial.
Deno Deploy: Free plan; Pro and Builder plans billed monthly with overage rates; enterprise custom with SLA.
Supabase Edge Functions: Free plan with an invocation allowance; Pro and Team tiers billed monthly; overage per million invocations; enterprise custom.
Netlify Functions: Free plan; Personal and Pro plans billed monthly in credits; function usage metered in credits; enterprise custom.
Vercel Functions: Free Hobby plan with a hard usage cap; Pro plan with included credit then pay-as-you-go rates; enterprise via sales.
Zeabur: Free plan; Dev, Pro and Team plans billed monthly; 14-day trial on Dev and Pro; enterprise by contract.
Cloudflare Workers: Free plan with a daily request limit; paid plan with a minimum charge plus request and CPU-time rates; enterprise by contract.
AWS Lambda: Published per-request and per-GB-second rates with a monthly free tier; other models for web functions and managed instances.
The other 4 quote per organisation.
Is there a free serverless platform?
Wasmer Edge, Val Town, Deno Deploy, Supabase Edge Functions, Netlify Functions, Vercel Functions, Zeabur, Cloudflare Workers, AWS Lambda have a free plan or free self-hosted edition.
Where are these serverless platform vendors established?
Vendor countries on this page, Elsewhere 10, North America 3, Asia-Pacific 1:
Wasmer Edge: Not stated by the vendor.
Val Town: Not stated by the vendor.
Bunny Edge Scripting: Not stated by the vendor.
Nuclio: Not stated by the vendor.
Deno Deploy: United States.
Koyeb: Not stated by the vendor.
Modal: Not stated by the vendor.
Supabase Edge Functions: Singapore.
Netlify Functions: United States.
Vercel Functions: Not stated by the vendor.
Zeabur: Not stated by the vendor.
Cloudflare Workers: United States.
AWS Lambda: Not stated by the vendor.
Azure Functions: Not stated by the vendor.
Which serverless platforms are open source?
Nuclio.
Which serverless platforms can you host yourself?
Nuclio. 13 products offer no self-hosted option.
What should you use instead of Wasmer Edge?
Second is Val Town, which suits Developers wiring up small cron jobs, webhooks and email handlers without managing infrastructure. Third is Bunny Edge Scripting, which suits teams adding TypeScript logic at the edge who want billing by requests and CPU.
Who should not buy Wasmer Edge?
Buyers who need a stated legal entity and country upfront. Active CPU time is capped at 6 hours a month on Hobby, which a busy API could exhaust..
We rank 14 serverless platforms products on this page. If we missed one, its maker can write to us about its purpose and customers. We pick what gets added; the terms of a listing are set out under software advice.
