Cloud & Data · Updated

Cloudflare Workers Pricing: $5 Minimum, Requests, CPU

Cloudflare Workers pricing combines a Free plan, a $5 monthly paid minimum, request charges, and CPU-time overages that shape serverless budgets.

AppStack Insider Editorial Team
AppStack Insider Editorial Team
AI-assisted research, human-reviewed • 7 min read
Cloudflare Workers Pricing: $5 Minimum, Requests, CPU

Cloudflare Workers pricing combines a default Free plan with a paid option that carries a $5 USD monthly minimum and usage-based billing. For B2B teams evaluating serverless workloads, the practical decision is not only how many requests an application serves, but also how much CPU time each invocation consumes.

What it is and who it’s for

Cloudflare Workers is Cloudflare’s serverless runtime for running application code at the edge, and its pricing page groups related services into the same paid plan framework. The Workers Paid plan includes Workers, Pages Functions, Workers KV, Hyperdrive, and Durable Objects usage, while D1 is available on both the Workers Free and Workers Paid plans.

That scope makes the offering relevant to application teams that are not buying only an execution environment. A buyer evaluating Cloudflare Workers pricing may also need to account for storage, database, queue, and stateful components that sit beside the Worker itself, including Workers KV, D1, Durable Objects, and Cloudflare Queues.

The Free plan is the default entry point. The Paid plan is the step for teams that need higher limits, broader feature access, or billing under the Standard usage model.

How it works

On the paid tier, billing runs on the Standard usage model. Cloudflare says changing the usage model affects only billable usage and has no technical implications.

For HTTP-triggered Workers, the main pricing levers are requests and CPU time. CPU time refers to the amount of processor time a Worker consumes during an invocation, which matters because the pricing page includes monthly CPU milliseconds and then bills additional usage beyond that allowance.

The limits page adds the operational context behind that billing model. Workers Free has a daily request limit, while Workers Paid has no request limit. Free HTTP requests get 10 milliseconds of CPU time per invocation, while Workers Paid can raise CPU time from a default of 30 seconds up to 5 minutes per invocation.

Execution duration and CPU time are not the same thing in the documentation. Cloudflare says HTTP-triggered Workers have no hard duration limit, while Cron Triggers, Durable Object Alarms, and Queue Consumers have a 15 minute duration limit.

The platform also prices some adjacent products with their own units. Cloudflare Queues charges are based on total operations against each queue during a month, and Durable Objects are billed for compute duration while actively running or idle in memory but unable to hibernate.

Pricing and cost considerations

Cloudflare Workers pricing has three components that matter for most paid deployments: a monthly minimum, included usage, and overage charges. The Workers Paid plan has a minimum charge of $5 USD per month for an account, and all included usage is on a monthly basis.

For the Standard usage model on paid Workers, Cloudflare documents the following core billing components for Worker execution:

ComponentWorkers Paid plan
Monthly minimum per account$5 USD
Included requests per month10 million
Additional requests$0.30 per additional million
Included CPU time per month30 million CPU milliseconds
Additional CPU time$0.02 per additional million CPU milliseconds

Cloudflare also states there are no additional charges for data transfer, also described as egress, or throughput, also described as bandwidth, on the Workers Paid plan.

The examples on Cloudflare’s pricing page show why request count alone does not explain the bill:

Example scenario from CloudflareResult
15 million requests/month at 7 ms CPU time per requestTotal $8.00
80% of 15 million requests handled without invoking the WorkerTotal $5.00
720 requests/month at 3 minutes of CPU time per requestTotal $6.99
100 million requests/month at 7 ms CPU time per requestTotal $45.40
100 million requests/month with 80 million requests cached and 20 million invoking the WorkerTotal $34.20

The examples cut both ways: a high share of cached or static-asset requests can keep a high-traffic, low-CPU workload near the $5 minimum, while a very low request count can still exceed the minimum when each invocation runs for minutes.

Feature-level pricing can also add cost outside core request and CPU billing. For example, Cloudflare Queues includes 1,000,000 operations per month and charges $0.40 per additional million operations, while Workers Logpush is only available on the Workers Paid plan and includes 10 million per month with additional usage billed at $0.05 per million.

Other included products have separate meters as well. Workers KV, Hyperdrive, D1, and Durable Objects each appear on the pricing page with their own included usage and overage structure, so total spend can rise even if Worker execution itself stays modest.

How to choose

The first decision point is whether the Free plan’s operational limits match the workload. Cloudflare documents a 100,000 requests per day limit on Workers Free, and exceeding that limit returns Error 1027.

The second decision point is whether the workload is CPU-light or CPU-heavy. Cloudflare says the average Worker uses approximately 2.2 ms per request, which suggests many request-response patterns may consume relatively little CPU compared with the paid plan’s monthly CPU allowance. That is a useful baseline for teams estimating whether request volume or compute time is more likely to drive spend. The paid plan’s usage-based-with-minimum structure is one of the patterns compared in our SaaS pricing models guide.

The third decision point is whether the application depends on higher platform limits rather than price alone. Paid accounts get 10,000 subrequests per invocation versus 50 on Free, 500 Workers per account versus 100, and 250 Cron Triggers per account versus 5. Teams weighing consumption-based platforms more broadly can compare this with credit-metered data infrastructure in our Snowflake pricing guide.

Limitations and gotchas

A common mistake is to treat Cloudflare Workers pricing as request-only billing. On the paid plan, CPU milliseconds are billed separately after included usage is exhausted, so two services with the same request count can produce different costs.

Another gotcha is that Free and Paid change limits well beyond billing. Memory per isolate stays at 128 MB on both plans, but the Worker size limit increases from 3 MB after compression on Free to 10 MB after compression on Paid, and environment variables increase from 64 per Worker to 128 per Worker.

Concurrency and integration limits can matter as much as headline price. Workers can have up to six simultaneous outgoing connections waiting for response headers, and the subrequest limit can be changed per Worker using the limits configuration.

Some buyers may also assume “serverless” means bandwidth fees apply on top. Cloudflare explicitly says there are no additional charges for egress or bandwidth on the Workers Paid plan.

Enterprise pricing is another boundary to note. Workers Enterprise accounts are billed based on the usage model specified in their contract.

FAQ

Does Workers Paid have a request-per-second cap?
Cloudflare’s limits page says there is no general limit on requests per second.

Do request body size limits come from the Workers plan?
No. Cloudflare says request body size limits depend on the Cloudflare account plan rather than the Workers plan.

Is there a response body size limit enforced by Workers?
Cloudflare says it does not enforce response body size limits, although CDN cache limits still apply.

How long are queue messages retained?
It depends on the plan. Cloudflare Queues retains messages for 24 hours, non-configurable, on Workers Free, and for 4 days by default on Workers Paid, configurable up to 14 days.

Sources

This article was produced with AI-assisted research and drafting and reviewed by a human editor. All sources are listed above. Read more about how we use AI and our editorial policy.

Spotted an inaccuracy? Email corrections@appstackinsider.com — see our corrections policy.

Related coverage

AppStack Insider Editorial Team

AppStack Insider Editorial Team

AI-assisted research, human-reviewed

AppStack Insider articles are produced with an AI-assisted research and drafting pipeline and reviewed by a human editor before publication. Every article cites its sources. See How We Use AI for the full process.

Don't miss the next market shift

Get our daily AI & SaaS insights delivered straight to your inbox.

By subscribing, you agree to our Privacy Policy.