We may earn a commission if you buy through links on this page. It never changes a score or a ranking. How we make money
Paid plans start at $250/mo. There's also a free plan.
Modal has a free Starter plan, but its first paid tier is Team at $250 per month, before usage-based compute charges. Starter includes $30 in monthly compute credits; Team includes $100. The plan fee buys capacity and account limits, not unlimited GPU time: CPU, memory and GPU use are metered, with H100 listed at $0.001097 per second. Modal lists no annual price. Enterprise pricing is custom.
01Every plan, line by line
| Plan | Billed annually | Billed monthly | Built for |
|---|---|---|---|
| Starter | — | $0 | Small teams and independent developers |
| Team | — | $250 | Startups and larger organizations |
| Enterprise | — | Quote | Organizations needing security and support |
Add-ons and overages
- Compute is charged by usage; CPU is $0.0000131/core/sec and memory is $0.00000222/GiB/sec.
- GPU rates vary by model; H100 is $0.001097/sec.
- Network egress is $0.04/GiB beyond plan allowances.
- Volumes include 1 TiB/month free, then cost $0.09/GiB/month.
- Region selection costs 1.15–1.75× base prices; non-preemptible execution costs 3× base prices.
- Shared Endpoint tokens are billed from the first request; token rates are not listed.
02What the plan grid doesn't tell you
Compute is metered
CPU costs $0.0000131/core/sec and memory costs $0.00000222/GiB/sec. GPU rates vary by model; H100 is $0.001097/sec.
Extra network egress
Egress beyond plan allowances costs $0.04/GiB. Starter includes 1 TiB/month free; Team includes 10 TiB/month.
Regions and execution
Region selection costs 1.15–1.75× base prices. Non-preemptible execution costs 3× base prices.
Volumes and endpoint tokens
Volumes include 1 TiB/month free, then cost $0.09/GiB/month. Shared Endpoint token rates are not listed; billing starts with the first request.
03Which plan should you buy?
| If you are… | Get | Why |
|---|---|---|
| Independent developers | Starter | It costs $0/month and includes $30 in monthly compute credits, but is capped at 3 seats and 10 GPU concurrency. |
| Startups and larger teams | Team | At $250/month, it includes $100 in compute credits, unlimited seats, 50 GPU concurrency and unlimited Scheduled Functions. |
| Organizations needing security | Enterprise | It offers custom compute allowance, higher GPU concurrency, audit logs, Okta SSO and HIPAA; pricing is not listed. |
04Is Modal worth the money?
Modal makes the most sense for teams that need inference, training, sandboxes or serverless compute in one usage-billed platform. The free Starter tier gives small teams room to start, but its 3-seat and 10-GPU-concurrency caps are real constraints. Team costs $250 monthly and includes $100 in compute credits, yet GPU and other compute usage can add to the bill. Buyers should model workload usage before treating the monthly fee as a predictable total.
Modal pricing FAQ
Does Modal have a free plan?
Yes. Starter is $0 per month and lists $30/month in compute credits, up to 3 seats, 100 containers, 10 GPU concurrency, 1 TiB monthly free network egress and 5 deployed crons. Compute is still usage-billed, so compare expected workloads with the included credits and rates.
How much does Modal Team cost?
Team is listed at $250 per month, with $100/month in compute credits, unlimited seats, 5,000 containers, 50 GPU concurrency, 10 TiB/month free egress and unlimited Scheduled Functions. The monthly fee does not make compute unlimited; Modal also charges for usage.
Does Modal offer annual billing?
The pricing page lists monthly plan fees and does not publish annual pricing. Starter is $0/month and Team is $250/month; Enterprise has no listed price and is custom. We would not assume an annual discount from the available plan grid.
How much does Modal GPU compute cost?
Modal does not list one flat GPU rate: rates vary by model. Its pricing lists H100 at $0.001097 per second. Region selection can cost 1.15–1.75× base prices, and non-preemptible execution is listed at 3× base prices, so workload configuration matters.