GPUBank vs Vast.ai
Vast.ai is a GPU marketplace from Los Angeles, founded in 2016, where independent hosts and vetted datacenters list machines and set their own prices. Its pricing page showed H100 SXM offers from $1.73 per GPU-hour on 2026-09-29, with a median of $2.13, for the GPU alone; GPUBank charges a fixed $1.85 with disk, IPv4 and traffic included. Vast.ai has 81 GPU cards, hosts in 52 countries and SOC 2 Type 2; GPUBank has fixed all-in prices and portable disks.
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
Vast.ai
Los Angeles, United States · since 2016
H100 SXM
$1.73 / h
- H100 SXM offers from $1.73 and RTX 4090 from $0.14 per GPU-hour
- 81 GPU cards on the pricing page, hosts in 52 countries
- Interruptible bids, with an H100 SXM minimum bid seen at $0.69
Vast.ai
Marketplace
United States28 points compared24 of their pages readprices last checked on 29 September 2026page updated 2 November 2026
The verdict
Lowest asks and the widest catalogue at Vast.ai. Fixed all-in prices at GPUBank.
Vast.ai suits price-sensitive users who compare offers and accept host variability: its cheapest live asks beat GPUBank on the H100 SXM, A100 SXM, L40S, RTX 5090 and RTX 4090, and its interruptible bids go lower still. Those are single cheapest offers, not list prices; the H100 SXM median was $2.13, the H200 median $4.80 and the B200 median $9.38. Vast.ai also bills disk and bandwidth separately at host-set rates, while GPUBank includes IPv4 and all traffic. Vast.ai has serverless endpoints, SOC 2 Type 2 and HIPAA BAAs on Secure Cloud; GPUBank has an SLA target and disks that move between regions.
Vast.aiGPUBank 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 | Vast.ai per hour | GPUBank per hour | Difference | 100 h at Vast.ai | 100 h at GPUBank |
|---|---|---|---|---|---|
 B200 SXM180 GB |
$6.25 | $4.15 | −34 % | $625.00 | $415.00 |
 H200 SXM141 GB |
$2.63 | $2.49 | −5 % | $263.00 | $249.00 |
 H100 SXM80 GB |
$1.73 | $1.85 | +7 % | $173.00 | $185.00 |
 H100 PCIe80 GB |
$1.93 | $1.35 | −30 % | $193.00 | $135.00 |
 RTX PRO 600096 GB |
$0.64 | $1.15 | +80 % | $64.00 | $115.00 |
 A100 SXM80 GB |
$0.31 | $0.95 | +206 % | $31.00 | $95.00 |
 L40S48 GB |
$0.47 | $0.55 | +17 % | $47.00 | $55.00 |
 RTX 509032 GB |
$0.33 | $0.48 | +45 % | $33.00 | $48.00 |
 RTX 409024 GB |
$0.14 | $0.23 | +64 % | $14.00 | $23.00 |
A fine-tune, 40 h on an H100 SXM
Vast.ai
$69.20
GPUBank
$74.00
Costs $4.80 more at GPUBank (+7 %)
An inference box, a month, 720 h, on an L40S
Vast.ai
$338.40
GPUBank
$396.00
Costs $57.60 more at GPUBank (+17 %)
Image generation, 200 h on an RTX 4090
Vast.ai
$28.00
GPUBank
$46.00
Costs $18.00 more at GPUBank (+64 %)
02Their catalogue
Every GPU Vast.ai lists.
Vast.ai lists 81 GPU models. The 62 below leave out its older and smaller cards, each at its lowest public price per GPU-hour. 9 of them are also rented at GPUBank (in blue).
- Blackwell and Hopper8
- B300288 GB · on demand$5.00
- B200192 GB · on demand$6.25
- H200141 GB · on demand$2.63
- H100 SXM80 GB · on demand$1.73
- H100 NVL80 GB · on demand$2.27
- H100 PCIe80 GB · on demand$1.93
- H200 NVL141 GB · on demand$2.20
- GB10128 GB · on demand$0.29
- Ampere, Ada and older data-centre12
- A100 SXM 80GB80 GB · on demand$0.31
A100 PCIe 80GB80 GB · on demand$0.42- A800 80GB80 GB · on demand$0.47
- L40S48 GB · on demand$0.47
- L4048 GB · on demand$0.34
L424 GB · on demand$0.16
A4048 GB · on demand$0.33
A1024 GB · on demand$0.20
T416 GB · on demand$0.15
V10016 GB · on demand$0.03
P4024 GB · on demand$0.09- P10016 GB · on demand$0.07
- Workstation18
- RTX PRO 600096 GB · on demand$0.64
RTX 6000 Ada48 GB · on demand$0.48
RTX A600048 GB · on demand$0.36
RTX A500024 GB · on demand$0.13
RTX 4000 Ada20 GB · on demand$0.19
RTX A400016 GB · on demand$0.07- RTX PRO 6000D96 GB · on demand$1.07
- RTX PRO 500032 GB · on demand$0.51
- RTX 5880 Ada48 GB · on demand$0.49
- RTX 4500 Ada24 GB · on demand$0.44
- RTX 5000 Ada32 GB · on demand$0.43
- RTX PRO 4500 Blackwell24 GB · on demand$0.28
Quadro RTX 800048 GB · on demand$0.25- RTX PRO 400020 GB · on demand$0.20
- Quadro RTX 600024 GB · on demand$0.13
- Quadro RTX 500016 GB · on demand$0.09
- Quadro P40008 GB · on demand$0.05
- RTX A20006 GB · on demand$0.03
- Consumer24
- RTX 509032 GB · on demand$0.33
- RTX 508016 GB · on demand$0.20
- Consumer continued24
- RTX 5070 Ti16 GB · on demand$0.12
- RTX 507012 GB · on demand$0.17
- RTX 5060 Ti16 GB · on demand$0.09
- RTX 50608 GB · on demand$0.08
- RTX 4090 D24 GB · on demand$0.67
- RTX 409024 GB · on demand$0.14
- RTX 4080 SUPER16 GB · on demand$0.21
- RTX 408016 GB · on demand$1.33
- RTX 4070 Ti SUPER16 GB · on demand$0.07
- RTX 4070 Ti12 GB · on demand$0.16
- RTX 4070 SUPER12 GB · on demand$0.08
- RTX 407012 GB · on demand$0.13
- RTX 4060 Ti8 GB · on demand$0.09
- RTX 40608 GB · on demand$0.06
RTX 3090 Ti24 GB · on demand$0.27- RTX 309024 GB · on demand$0.12
RTX 3080 Ti12 GB · on demand$0.11- RTX 308010 GB · on demand$0.08
- RTX 3070 Ti8 GB · on demand$0.11
- RTX 30708 GB · on demand$0.07
- RTX 3060 Ti8 GB · on demand$0.07
- RTX 306012 GB · on demand$0.03
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.
Vast.ai
2 h 14 m 03 s on H100 SXM
- GPU · $1.73 / h
- $3.87
- vCPU and RAM
- Included
- Storage
- Host-set, $0.0013 to $1.00 per GB-month
- Egress
- Metered both ways, host-set, up to $68.27/TB
- Public IPv4
- Shared host IP with mapped ports by default
- Billing granularity
- Per second
- Minimum charge
- None, no rounding up
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.
Vast.ai
GPU released; disk billed at host rate.
Docs note that stopped-instance storage rates can be higher than running rates.
GPUBank
Only the persistent disk, $0.03 per GB-month. Compute, IPv4 and scratch cost nothing.
04Where the GPUs run
52 countries against 4, and where they are.
Countries where Vast.ai 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
Vast.ai52
- United States
Canada
Czech Republic
South Korea
Taiwan
United Kingdom
Bulgaria
Japan
Romania
Germany
Spain
South Africa
Poland
China
Belgium
Sweden
Argentina
Ireland
Ukraine
Finland
Netherlands
France
Estonia
Slovakia
Norway
Saudi Arabia
Denmark
Thailand
India
Australia
Russia
Hungary
Slovenia
Italy
Croatia
Brazil
Singapore
Vietnam
Turkey
Iceland
Greece
Switzerland
Israel
Oman
Peru
Indonesia
Latvia
Austria
Hong Kong
United Arab Emirates
Mexico
Sri Lanka
GPUBank4
- Paris
- Frankfurt
- Ashburn
- Singapore
Vast.ai · 52 countriesGPUBank · 4 regions
05Paying and signing up
How each one takes your money.
Vast.ai
Payments and sign-up
Payment methods
Cards
CryptoCrypto: BitPay and Crypto.com
Minimum top-up $5
To start
- 1Email and password
- 2Email verification
Billing modelPrepaid credits
GPUBank
Payments and sign-up
Payment methods
No cardsA 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.
| Vast.ai | GPUBank | |
|---|---|---|
| SOC 2 | Type II | Not published |
| ISO 27001 | Not published | Not published |
| HIPAA | BAA available | No |
| GDPR DPA | DPA available | Published DPA, applies automatically |
| Availability target | None for standard instances | 99.9 % monthly, per region |
| Status page | status.vast.ai | Public status page |
| Support | Chat, Discord, contact form | 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 Vast.ai pricing page captured on 29 September 2026, and GPUBank's live board. Cookie banners and chat bubbles are closed; nothing else is edited.
- 1The pricing page lists 81 GPU cards in 11 architecture groups, each with a from price and a median, without login.
- 2Prices update hourly; two cards show only an indicative price marked No current offers.
- 3The reserved tier leads to Contact Sales, with terms of 1, 3 or 6 months and up to 50% off.
- 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 aheadVast.ai aheadEven
| Criterion | GPUBank | Vast.ai |
|---|---|---|
| Price and catalogue | ||
| H100 SXM, per GPU-hourCheapest live offer on 2026-09-29, GPU rate only. Disk and bandwidth are billed on top. | $1.85, all-in | From $1.73 lowest ask; median $2.13Ahead |
| Cheapest GPU-hour | $0.23, RTX 4090 | $0.03 lowest ask, GTX 1060 and othersAhead |
| GPU models listed | 10 | 81Ahead |
| What the GPU price includes | vCPU, RAM, NVMe, IPv4, traffic | vCPU and RAM; disk and bandwidth extra |
| Billing | ||
| Billing granularity | Per second | Per second |
| Minimum charge | None | None, no rounding up |
| Stopped instanceDocs note that stopped-instance storage rates can be higher than running rates. | Disk only | GPU released; disk billed at host rate |
| Persistent storageRange and median of 0.20 USD/GB-month from sampled offers in the public API. | $0.03 per GB-month | Host-set, $0.0013 to $1.00 per GB-month |
| Egress trafficMedian of 2.67 USD/TB in sampled offers; not shown in the hourly price. | Included | Metered both ways, host-set, up to $68.27/TB |
| Public IPv4Some hosts list static IPs. | Included | Shared host IP with mapped ports by default |
| Billing model | Prepaid balance | Prepaid credits |
| Payment and sign-up | ||
| Crypto paymentsCredits bought with crypto are not refundable. | Bitcoin, USDT, Ethereum and 12+ more | BitPay and Crypto.com |
| Card payments | No | Cards via StripeAhead |
| Minimum deposit | $25 | $5Ahead |
| Platform | ||
| API and CLI | REST API, CLI, Python SDK | REST API, CLI, Python SDK |
| Terraform | Official provider | Not in the docs index |
| Serverless GPUs | No | Yes, at the instance rateAhead |
| Managed Kubernetes | No | No |
| Ready-made templates | 25 | Not published |
| Interruptible or spotLowest H100 SXM minimum bid seen in the public API: $0.69 per GPU-hour. | Yes, 40 % off | Interruptible bidsAhead |
| GPUs per instance | Up to 8 | Up to 8 per offer; clusters via salesAhead |
| Multi-node interconnectNVLink presence and bandwidth vary by host offer. | Not offered | InfiniBand on sales-led clustersAhead |
| Network and regions | ||
| Regions | 4 | No named regions; host locationsAhead |
| Countries | France, Germany, US, Singapore | 52 in sampled offersAhead |
| EU regions | Paris, Frankfurt | Hosts in 22 EU countries in sampled offersAhead |
| Trust and support | ||
| Availability target | 99.9 % monthly | None for standard instances |
| Compliance reportsThe SOC 2 report is shared under NDA. | None published | SOC 2 Type 2, HIPAA BAAs on Secure CloudAhead |
| Public status page | Yes | status.vast.ai |
| SupportThe clusters page mentions 24/7 support for cluster customers. | Tickets from a $25 balance | Chat, Discord, contact formAhead |
09The right fit
Different needs. A clear choice.
What Vast.ai does better
- H100 SXM offers from $1.73 and RTX 4090 from $0.14 per GPU-hour
- 81 GPU cards on the pricing page, hosts in 52 countries
- Interruptible bids, with an H100 SXM minimum bid seen at $0.69
- Serverless endpoints and a public offer search API
- SOC 2 Type 2, HIPAA BAAs on Secure Cloud, $5 minimum deposit
Choose GPUBank if
- Fixed prices, where Vast.ai asks are host-set and change hourly
- IPv4 and all traffic included; Vast.ai bills bandwidth up to $68.27/TB
- B200 at $4.15 and H200 at $2.49, against offers from $6.25 and $2.63
- Stop an instance for $0.03 per GB-month and keep a movable disk
- MI300X and a 99.9 % monthly SLA target
Moving from Vast.ai takes an afternoon.
Vast.ai data sits on host-local disks and volumes that cannot move between machines, so it must be copied out before the instance is destroyed. Workloads are standard Docker images and move as they are.
- 1Launch the same GPU on GPUBank with a persistent disk in the region you need.
- 2Copy data with scp or rsync, or through Vast.ai Cloud Sync to S3, then pull it onto the GPUBank disk.
- 3Run the same Docker image or pick the matching template, then destroy the Vast.ai instances.
# 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.
At the cheapest offer, yes. Vast.ai showed H100 SXM offers from $1.73 per GPU-hour on 2026-09-29, with a median of $2.13, and disk and bandwidth are billed on top. GPUBank charges a fixed $1.85 per GPU-hour with vCPU, RAM, NVMe, IPv4 and traffic included.
No. Vast.ai is a marketplace: each host sets its own price, and the pricing page updates hourly. The from price is a single cheapest offer. Storage and bandwidth rates are also set per host, with internet cost up to $68.27 per TB in sampled offers.
The GPU charge stops and the GPU is released to other renters. The container disk keeps billing at the host's storage rate until the instance is destroyed. On GPUBank a stopped instance costs only its persistent disk, at $0.03 per GB-month, and the disk can move between regions.
Yes, through BitPay and Crypto.com, and cards through Stripe, from a $5 deposit. Credits bought with crypto cannot be refunded. GPUBank takes Bitcoin, USDT, Ethereum and more than a dozen other coins from $25, and no cards.
Its compliance page states SOC 2 Type 2, with the report under NDA, a SOC 3 report on request and HIPAA BAAs on the Secure Cloud tier. Community hosts hold no formal certification. GPUBank publishes no SOC 2 or ISO 27001 report.
Without. Every price on this page excludes VAT and sales tax, for Vast.ai 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 Vast.ai was read on its own pages or official documentation. Prices last checked on 29 September 2026.
- 01vast.ai/pricing81 GPU cards with from/median prices, instance types, per-second billing, no minimum
- 02vast.ai/Scale claims, $5 start, CLI/SDK/API, serverless, clusters
- 03vast.ai/pricing/gpu/H100-SXMConfirms per-GPU hourly 'from' price and hourly update
- 04console.vast.ai/api/v0/search/asks/Interruptible minimum bids, storage and bandwidth rate ranges, host countries
- 05docs.vast.ai/guides/instances/pricingCost components, reserved up to 50%, interruptible 50%+, credits upfront
- 06docs.vast.ai/guides/instances/choosing/instance-typesOn-demand, reserved, interruptible behaviour
- 07docs.vast.ai/guides/reference/faq/rental-typesInterruptible bidding, 50-80% savings
- 08docs.vast.ai/guides/instances/choosing/reserved-instancesTerms, refunds, 5% cancellation fee, no host migration
- 09docs.vast.ai/guides/reference/billingPrepaid credits, Stripe cards, BitPay and Crypto.com, refunds, storage and bandwidth billing
- 10docs.vast.ai/guides/reference/faq/billingCard and email verification for card billing, no PayPal
- 11docs.vast.ai/guides/get-started/quickstartSign-up steps, email verification, 5 USD minimum deposit
- 12docs.vast.ai/guides/instances/storage/typesContainer storage and volumes, stopped-instance storage billing
- 13docs.vast.ai/guides/instances/storage/volumesLocal volume limits
- 14docs.vast.ai/guides/reference/faq/networkingPort mapping on shared public IP
- 15docs.vast.ai/guides/serverless/pricingServerless billed at instance rates per worker state
- 16docs.vast.ai/guides/instances/virtual-machinesVM support on a subset of machines
- 17docs.vast.ai/guides/reference/faq/securitySecure Cloud definition, isolation
- 18docs.vast.ai/guides/reference/faq/generalLinux Docker only, SSH/Jupyter, Secure Cloud filter
What we could not verify (9)
- Reserved (1/3/6 month) per-GPU prices are not published; only 'up to 50%' and a worked example (on-demand 720 USD/month to 576 USD/month for 1 month) in docs
- Secure Cloud price levels per model are not published separately on the pricing page
- Interruptible prices are a live API snapshot (lowest min_bid among up to 64 returned offers; the unauthenticated API returns at most 64 offers per query), not a published list
- Terraform provider and Kubernetes support: not found in the official docs index
- Number of templates not published as a figure
- Snapshot feature not documented in the pages read
- Funding and investors not published on vast.ai
- Named regions do not exist; country list derived from host geolocation in the sampled API offers, not an official region list
- SLA for standard instances not published; reserved docs refer to an 'implicit or explicit Service Level Agreement' per machine without figures
In short
GPUBank vs Vast.ai, 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
- Vast.ai: From $1.73 lowest ask; median $2.13
GPUBank: $1.85, all-in - Cheapest GPU-hour
- Vast.ai: $0.03 lowest ask, GTX 1060 and others
GPUBank: $0.23, RTX 4090 - Billing granularity
- Vast.ai: Per second
GPUBank: Per second - Stopped instance
- Vast.ai: GPU released; disk billed at host rate
GPUBank: Disk only - Persistent storage
- Vast.ai: Host-set, $0.0013 to $1.00 per GB-month
GPUBank: $0.03 per GB-month - Egress traffic
- Vast.ai: Metered both ways, host-set, up to $68.27/TB
GPUBank: Included - Crypto payments
- Vast.ai: BitPay and Crypto.com
GPUBank: Bitcoin, USDT, Ethereum and 12+ more - Card payments
- Vast.ai: Cards via Stripe
GPUBank: No - Minimum deposit
- Vast.ai: $5
GPUBank: $25 - GPU models listed
- Vast.ai: 81
GPUBank: 10 - Regions
- Vast.ai: No named regions; host locations
GPUBank: 4 - Serverless GPUs
- Vast.ai: Yes, at the instance rate
GPUBank: No - Availability target
- Vast.ai: None for standard instances
GPUBank: 99.9 % monthly - Compliance reports
- Vast.ai: SOC 2 Type 2, HIPAA BAAs on Secure Cloud
GPUBank: None published - Support
- Vast.ai: Chat, Discord, contact form
GPUBank: Tickets from a $25 balance - Headquarters
- Vast.ai: Los Angeles, United States
GPUBank: Tallinn, Estonia
How this page was made
Every price and policy of Vast.ai 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. Vast.ai, its logo and its website belong to their owner and are shown only to identify the service compared. GPUBank is not affiliated with Vast.ai.
