RunPod vs Modal - $0 vs $30 per month free credit

RunPod is an AI developer GPU cloud that sells Pods, Serverless, and Clusters, billed per second. Modal is a serverless compute platform where you deploy Python functions that can attach GPUs, billed at per-second rates. All figures below come from each vendor’s official pricing and billing docs, checked on 2026-08-20.

Billing granularity

RunPod’s billing docs and Pods pricing page say compute and storage are billed per second, with no data-transfer fees; serverless is billed from when a worker starts until it fully stops, rounded up to the nearest second. There is a wrinkle, though. A third official page, Pods/overview, says “Pods are billed by the minute,” and the Cloud GPU product FAQ mixes “by the millisecond” with “billed by the second.” For a billing claim, prefer the billing and Pods pricing pages, and do not treat “per second” as the only official sentence RunPod publishes.

Modal’s pricing page displays GPU, CPU, and memory rates per second, and its billing docs say there are “no minimum usage-time increments.” Note the docs phrase it that way rather than using the exact words “billed per second.” So both land on per-second granularity, but RunPod’s own pages disagree with each other on Pods, while Modal’s docs are internally consistent.

GPU range and pricing

RunPod’s on-demand Pods span a wide range. The cheapest listed GPU is an RTX A5000 at $0.16/hr on Community Cloud ($0.27/hr on Secure Cloud), and the most expensive is a B300 at $6.94/hr Community / $7.89/hr Secure. RunPod Serverless rows run from $0.58/hr for the A4000/A4500/RTX 4000/RTX 2000 tier up to $9.98/hr for a B300. The official split is Secure Cloud versus Community Cloud, not “consumer versus datacenter.”

Modal prices GPU per second. Its pricing page lists GPU, CPU, memory, and volumes as separate per-second line items; the verified facts record one concrete GPU rate, an Nvidia H100 at $0.001097/sec, which I cite rather than a full Modal range. Modal’s Starter plan ($0 + compute/month) includes $30/month of “included compute,” which offsets usage costs. Whether that $30 specifically covers GPU is not stated officially, so I will not call it free GPU credit.

Screenshot: RunPod pods pricing page

Last checked: 2026-08-20

Screenshot: Modal pricing page

Last checked: 2026-08-20

Product shape

RunPod sells three shapes. Pods are dedicated GPU instances you spin up and use like a box; Serverless exposes API inference endpoints; Clusters run multi-node GPU jobs across nodes. Modal is purely serverless: you deploy Python functions that can attach GPUs, and there is no persistent SSH box to administer. If your workflow assumes a long-running machine you log into and configure, that is RunPod Pods, not Modal.

Idle costs

RunPod Pods keep costing you when stopped, but only for storage, not compute. The docs say “storage charges continue to accrue on stopped Pods”: a volume disk is $0.10/GB/mo while running and $0.20/GB/mo while idle, and it persists until you delete the Pod. Container disk, by contrast, is temporary and is not charged when the Pod stops. RunPod Serverless Flex scales to zero when idle, with a default idle timeout of 5 seconds after the last request.

Modal is serverless end to end, so idle compute scales to zero: you pay only for the compute you use or request, with no minimum usage-time increments. There is no stopped-but-still-billed storage box to forget about, the way a RunPod Pod can quietly accrue volume charges while it sits idle.

Payment and onboarding

RunPod has no advertised free signup credit. Its billing docs suggest you “can start with as little as $10” deposited, but that is a minimum deposit suggestion, not a grant, so do not read it as free money. Modal requires a payment method on file to use the platform at all, per its billing docs.

How to choose

Choose RunPod if you want a persistent GPU box you administer, you run multi-node cluster jobs, or you need a wide on-demand GPU range with Community Cloud pricing down to $0.16/hr. Keep an eye on stopped-Pod volume charges, and on the per-second-versus-per-minute wording scattered across RunPod’s own pages.

Choose Modal if your workload is bursty and should scale to zero when idle, you prefer deploying functions over managing servers, or you want a monthly $30 included-compute allowance on a $0 Starter plan. Remember a payment method is required to use Modal.

For more on Modal’s free tier, see the Modal free tier credit entry, and for the broader decision, read how to choose between a VPS, GPU cloud, serverless, or a sandbox. Every grant I track is on the credits index.