GPUBank vs Modal
Modal is a US serverless platform: you define functions, images and GPU needs in Python, and it runs containers that scale to zero between requests. Its H100 SXM costs $0.001097 per second, or $3.95 per hour, for the GPU alone; CPU and RAM are billed on top per second. GPUBank rents a full H100 SXM VM at $1.85 per GPU-hour with vCPU, RAM, NVMe and traffic included.
GPUBank
Tallinn, Estonia
H100 SXM
$1.85 / h
- vCPU, RAM, NVMe, IPv4 and traffic included
- Per-second billing after the health check
- Top up in Bitcoin, USDT and 12+ other coins
vCPU, RAM, NVMe, IPv4 and traffic included
Modal
New York, United States
H100 SXM · serverless
$3.95 / h
- Serverless containers that scale to zero, billed per second
- $30 of free compute each month on the free Starter plan
- B300, A100 40GB, A10, L4 and T4 in the catalogue
Modal
Serverless
United States28 points compared12 of their pages readprices last checked on 29 September 2026page updated 2 November 2026
The verdict
Scale-to-zero serverless at Modal. Lower all-in prices and full VMs at GPUBank.
Modal fits spiky inference and batch jobs: you pay only while containers run, and the free Starter plan includes $30 of compute each month. GPUBank fits long-running work on a VM with SSH, where the GPU price already covers vCPU, RAM, NVMe and traffic. On the five GPUs both of us list, Modal's GPU-only rate is 34 to 72 % higher, before CPU, RAM and a 1.15x to 1.75x multiplier if you pin a region. Every Modal GPU function is preemptible; a GPUBank instance is not, unless you pick interruptible.
ModalGPUBank 01The real cost
The same GPUs, side by side, per hour.
Every GPU model both of us rent, at each provider's own published price for one card, in US dollars, excluding VAT. Prices last checked on 29 September 2026.
| GPU | Modal per hour | GPUBank per hour | Difference | 100 h at Modal | 100 h at GPUBank |
|---|---|---|---|---|---|
 B200 SXM180 GB · their serverless price |
$6.25 | $4.15 | −34 % | $625.00 | $415.00 |
 H200 SXM141 GB · their serverless price |
$4.54 | $2.49 | −45 % | $454.00 | $249.00 |
 H100 SXM80 GB · their serverless price |
$3.95 | $1.85 | −53 % | $395.00 | $185.00 |
 RTX PRO 600096 GB · their serverless price |
$3.03 | $1.15 | −62 % | $303.00 | $115.00 |
 L40S48 GB · their serverless price |
$1.95 | $0.55 | −72 % | $195.00 | $55.00 |
Where Modal publishes no on-demand price for a model, its lowest published serverless price is used, as labelled on the row.
A fine-tune, 40 h on an H100 SXM
Modal
$158.00
GPUBank
$74.00
Saves $84.00 at GPUBank (−53 %)
An inference box, a month, 720 h, on an L40S
Modal
$1,404.00
GPUBank
$396.00
Saves $1,008.00 at GPUBank (−72 %)
A long-context job, 40 h on an H200
Modal
$181.60
GPUBank
$99.60
Saves $82.00 at GPUBank (−45 %)
02Their catalogue
Every GPU Modal lists.
11 GPU models listed by Modal, each at its lowest public price per GPU-hour. 5 of them are also rented at GPUBank (in blue).
- B300serverless$7.10
- B200serverless$6.25
- H200141 GB · serverless$4.54
- H100 SXMserverless$3.95
- A100 40GB40 GB · serverless$2.10
- A100 80GB80 GB · serverless$2.50
- L40S48 GB · serverless$1.95
L4serverless$0.80
A10serverless$1.10
T4serverless$0.59- RTX PRO 6000serverless$3.03
03What the price includes
What appears on the bill besides the GPU.
The same H100 SXM session of 2 h 14 m 03 s, line by line, as each provider describes its billing.
Modal
2 h 14 m 03 s on H100 SXM
- GPU · $3.95 / h
- $8.82
- vCPU and RAM
- Billed separately
- Storage
- $0.09 per GiB-month, 1 TiB included
- Egress
- $0.04 per GiB over allowance, from 1 Oct 2026
- Public IPv4
- Shared NAT; static IP proxy on Team plan
- Billing granularity
- Per second
- Minimum charge
- None; 0.125 core minimum per container
GPUBank
2 h 14 m 03 s on H100 SXM
- GPU · $1.85 / h
- $4.13
- vCPU and RAM
- Included
- Storage
- Local NVMe included; persistent disk $0.03 per GB-month
- Egress
- Included
- Public IPv4
- Included
- Billing granularity
- Per second, from the health check
- Minimum charge
- None
While stopped
What keeps running on the bill.
Modal
Serverless, scales to zero.
No VM to stop. Containers are billed while alive.
GPUBank
Only the persistent disk, $0.03 per GB-month. Compute, IPv4 and scratch cost nothing.
04Where the GPUs run
8 countries against 4, and where they are.
Countries where Modal says its GPUs run, from its own region list or documentation, against GPUBank's four regions. A pin marks a country, not the exact site.
Paris
Frankfurt
Ashburn
Singapore
Modal8
- United States
Japan
Australia
United Kingdom
Canada
Mexico
Ireland
India
GPUBank4
Paris
Frankfurt- Ashburn
Singapore
Modal · 8 countriesGPUBank · 4 regions
05Paying and signing up
How each one takes your money.
Modal
Payments and sign-up
Payment methods
Cards
Invoice, enterprise only
No cryptoMinimum top-up Not published
To start
- 1Email
- 2Card
Billing modelPostpaid monthly, card on file
GPUBank
Payments and sign-up
Payment methods
A prepaid USD balance, topped up in crypto. Every GPU-second is drawn from it.
Minimum top-up $25
To start
- 1Email and password
- 2Top up in crypto
- 3Launch a GPU
Billing modelPrepaid balance
06Trust and support
Certificates, uptime and who answers.
| Modal | GPUBank | |
|---|---|---|
| SOC 2 | Type II | Not published |
| ISO 27001 | Not published | Not published |
| HIPAA | BAA available | No |
| GDPR DPA | Not published | Published DPA, applies automatically |
| Availability target | No SLA found | 99.9 % monthly, per region |
| Status page | status.modal.com | Public status page |
| Support | Community Slack; private Slack on Enterprise | Tickets from a $25 balance |
07Their site, our site
What you see before you pay.
Both sites as a visitor sees them, without an account: the Modal pricing page captured on 29 September 2026, and GPUBank's live board. Cookie banners and chat bubbles are closed; nothing else is edited.
- 1GPU prices are public per second and per hour, with a toggle between the two.
- 2The plan table shows the 1.15x to 1.75x region multipliers and a 3x multiplier for non-preemptible CPU work.
- 3FAQ answers are not rendered in the page HTML.
- 1Live stock per region and every price, without an account.
- 2Per-hour and per-second prices, with vCPU, RAM, NVMe, IPv4 and traffic in them.
- 3A machine-readable price feed at /prices.json.
08The spec sheet
28 points, read off both sites.
GPUBank aheadModal aheadEven
| Criterion | GPUBank | Modal |
|---|---|---|
| Price and catalogue | ||
| H100 SXM, per GPU-hourPublished as $0.001097 per second. CPU and RAM are billed on top. | $1.85, all-in | $3.95 per hour, GPU only |
| Cheapest GPU-hour | $0.23, RTX 4090 | $0.59, T4, GPU only |
| GPU models listed | 10 | 11Ahead |
| What the GPU price includes | vCPU, RAM, NVMe, IPv4, traffic | GPU only; CPU and RAM billed per second |
| Billing | ||
| Billing granularity | Per second | Per second |
| Minimum charge | None | None; 0.125 core minimum per container |
| Stopped instanceNo VM to stop. Containers are billed while alive. | Disk only | Serverless, scales to zero |
| Persistent storageThe first 1 TiB of Volumes each month is free; above it Modal costs more per GiB. | $0.03 per GB-month | $0.09 per GiB-month, 1 TiB included |
| Egress trafficAllowance per billing cycle: 1 TiB on Starter, 10 TiB on Team, 100 TiB on Enterprise. | Included | $0.04 per GiB over allowance, from 1 Oct 2026 |
| Public IPv4 | Included | Shared NAT; static IP proxy on Team plan |
| Billing model | Prepaid balance | Postpaid monthly, card on file |
| Payment and sign-up | ||
| Crypto payments | Bitcoin, USDT, Ethereum and 12+ more | Not listed |
| Card payments | No | Cards via StripeAhead |
| Minimum deposit | $25 | Not published |
| Platform | ||
| API and CLI | REST API, CLI, Python SDK | Python SDK and CLI; JS and Go SDKs in beta |
| Terraform | Official provider | None documented |
| Serverless GPUs | No | Yes, the core productAhead |
| Managed Kubernetes | No | No |
| Ready-made templates | 25 | Examples library |
| Interruptible or spotGPUBank offers both standard instances and interruptible at -40 %. | Yes, 40 % off | All GPU functions preemptible, no discount |
| GPUs per instance | Up to 8 | Up to 8, clusters up to 256Ahead |
| Multi-node interconnect | Not offered | RDMA 3,200 Gb/s in multi-node clustersAhead |
| Network and regions | ||
| RegionsPinning a region multiplies all usage by 1.15x to 1.75x. | 4 | US, EU, APAC and 8 more areasAhead |
| Countries | France, Germany, US, Singapore | 8, incl. US, UK, Canada, Japan, AustraliaAhead |
| EU regions | Paris, Frankfurt | eu-west, eu-north, eu-southAhead |
| Trust and support | ||
| Availability target | 99.9 % monthly | No SLA found |
| Compliance reports | None published | SOC 2 Type II; HIPAA BAA on EnterpriseAhead |
| Public status page | Yes | status.modal.com |
| Support | Tickets from a $25 balance | Community Slack; private Slack on Enterprise |
09The right fit
Different needs. A clear choice.
What Modal does better
- Serverless containers that scale to zero, billed per second
- $30 of free compute each month on the free Starter plan
- B300, A100 40GB, A10, L4 and T4 in the catalogue
- Multi-node clusters up to 256 GPUs with 3,200 Gb/s RDMA
- SOC 2 Type II, and a HIPAA BAA on the Enterprise plan
Choose GPUBank if
- H100 SXM at $1.85 all-in, against $3.95 for Modal's GPU alone
- vCPU, RAM, NVMe, IPv4 and all traffic included in the GPU price
- Instances are not preempted unless you choose interruptible
- Full VMs with SSH, JupyterLab and Terraform, no SDK rewrite
- Pay in Bitcoin, USDT and 12+ other coins, no card
Moving from Modal takes an afternoon.
Modal code is written against its Python SDK, so moving to a VM means replacing the decorators with your own entry point. Container images are standard, and Volume data can be downloaded with the Modal CLI; egress above the allowance is billed.
- 1Launch the same GPU on GPUBank with a persistent disk in the region you need.
- 2Download your Volumes with the Modal CLI and copy them to the disk with rsync.
- 3Run the same container image with a plain entry point or a web server in place of the Modal decorators.
# 1. a GPU and a disk on GPUBank gpubank launch --gpu h100-sxm --region eu-paris --disk data # 2. copy your data from the old machine rsync -azP old-host:/workspace/ root@<new-ip>:/workspace/ # 3. run the same image or template ssh root@<new-ip> 'cd /workspace && ./start.sh'
A few things you might be wondering.
Yes, for sustained use. Modal charges $0.001097 per second for an H100 SXM, $3.95 per hour, for the GPU alone, then bills CPU and RAM on top. GPUBank charges $1.85 per GPU-hour with vCPU, RAM, NVMe, IPv4 and traffic included.
When traffic is spiky. Modal scales containers to zero, so idle time costs nothing, and the free Starter plan includes $30 of compute a month. A GPUBank instance bills every second it runs, even when idle.
Yes. Modal's docs say GPU functions are always preemptible and cannot be made non-preemptible. GPUBank instances run until you stop them, unless you pick the interruptible option at -40 %.
Modal does not document crypto or PayPal. A card on file is required to use GPUs, and invoices with bank transfer are for Enterprise plans with a commitment. GPUBank is paid from a USD balance topped up in Bitcoin, USDT, Ethereum and more than a dozen other coins.
Modal does not offer an SSH-accessible VM; work runs as functions, sandboxes or notebooks in its container model. GPUBank gives you a VM with SSH, JupyterLab and custom Docker images.
Without. Every price on this page excludes VAT and sales tax, for Modal and for GPUBank, so the two columns compare like for like. Tax is added at checkout according to where the buyer is based.
Sources
Every figure about Modal was read on its own pages or official documentation. Prices last checked on 29 September 2026.
- 01modal.com/pricingGPU per-second and per-hour rates, CPU/memory, volumes, plans, region and non-preemptible multipliers
- 02modal.com/docs/guide/gpuGPU list, max GPUs per container, SXM, auto-upgrades, payment method required
- 03modal.com/docs/guide/billingMonthly postpaid billing, Stripe, Enterprise invoicing and bank transfer
- 04modal.com/docs/guide/network-egress-billingEgress pricing from 1 Oct 2026
- 05modal.com/docs/guide/resourcesDefault requests, request vs usage billing, disk
- 06modal.com/docs/guide/region-selectionRegions, multipliers, routing regions
- 07modal.com/docs/guide/preemptionGPU functions always preemptible
- 08modal.com/docs/guide/multi-node-clustersRDMA bandwidth, 256 devices
- 09modal.com/docs/guide/securitySOC 2 Type II, HIPAA BAA, Stripe
- 10modal.com/companyFunding, offices
- 11status.modal.com/Status page uptime
- 12sec.gov/Archives/edgar/data/2088132/000208813226000002/xslFormDX01/primary_doc.xmlHeadquarters
What we could not verify (6)
- A100 80GB form factor (SXM or PCIe) not published
- VRAM for B300, B200, H100, RTX PRO 6000, A10, L4, T4 not stated on pages read
- Formal SLA not found
- Founding year, legal entity and HQ not stated on pages read
- Crypto and PayPal not documented
- Volume discounts and committed rates are Enterprise contact-sales
In short
GPUBank vs Modal, in facts.
The essentials of this comparison in one table, for quick readers and AI assistants. Prices last checked on 29 September 2026.
Read as Markdown- H100 SXM, per GPU-hour
- Modal: $3.95 per hour, GPU only
GPUBank: $1.85, all-in - Cheapest GPU-hour
- Modal: $0.59, T4, GPU only
GPUBank: $0.23, RTX 4090 - Billing granularity
- Modal: Per second
GPUBank: Per second - Stopped instance
- Modal: Serverless, scales to zero
GPUBank: Disk only - Persistent storage
- Modal: $0.09 per GiB-month, 1 TiB included
GPUBank: $0.03 per GB-month - Egress traffic
- Modal: $0.04 per GiB over allowance, from 1 Oct 2026
GPUBank: Included - Crypto payments
- Modal: Not listed
GPUBank: Bitcoin, USDT, Ethereum and 12+ more - Card payments
- Modal: Cards via Stripe
GPUBank: No - Minimum deposit
- Modal: Not published
GPUBank: $25 - GPU models listed
- Modal: 11
GPUBank: 10 - Regions
- Modal: US, EU, APAC and 8 more areas
GPUBank: 4 - Serverless GPUs
- Modal: Yes, the core product
GPUBank: No - Availability target
- Modal: No SLA found
GPUBank: 99.9 % monthly - Compliance reports
- Modal: SOC 2 Type II; HIPAA BAA on Enterprise
GPUBank: None published - Support
- Modal: Community Slack; private Slack on Enterprise
GPUBank: Tickets from a $25 balance - Headquarters
- Modal: New York, United States
GPUBank: Tallinn, Estonia
How this page was made
Every price and policy of Modal was read on its own website, documentation or public API on 29 September 2026, with the address of each page listed above. A value we could not read on a primary source is marked "Not published" rather than estimated. GPUBank figures come from this site and its terms. Prices change: check both providers before you buy. Modal, its logo and its website belong to their owner and are shown only to identify the service compared. GPUBank is not affiliated with Modal.
