GPUBank vs Together AI
Together AI is a San Francisco AI cloud built around per-token inference on open-weight models, which also sells self-serve GPU clusters of H100, H200, B200 and B300 in multiples of 8 GPUs. Its on-demand H100 SXM costs $3.99 per GPU-hour on a cluster; GPUBank charges $1.85 for a single GPU. The $3.99 dedicated inference H100 is a promotion valid until 30 Sep 2026, against a list price of $5.49. Together wins on cluster scale, inference products and certifications; GPUBank wins on price, unit size, billing and crypto payment.
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
Together AI
San Francisco, United States
H100 SXM
$3.99 / h
- Self-serve InfiniBand clusters of 8 to 256 GPUs, 4,000+ reserved
- B300 on demand at $9.99, GB200 and GB300 NVL72 through sales
- Serverless inference, dedicated endpoints and fine-tuning on one account
Together AI
GPU cloud
United States25 points compared24 of their pages readprices last checked on 29 September 2026page updated 2 November 2026
The verdict
Clusters, inference and certifications at Together AI. About half the price and single GPUs at GPUBank.
Together AI suits teams that train on 8 to 256 Hopper or Blackwell GPUs with managed Kubernetes or Slurm, or that already use its inference API, and that need SOC 2 Type II and ISO 27001 paperwork. For the three GPUs both of us rent, GPUBank costs 49 to 58 % less per GPU-hour on demand. Together clusters start at 8 GPUs and bill hourly until deleted; GPUBank rents one GPU, bills by the second and keeps only the disk on the bill when you stop. Together takes cards and US ACH; GPUBank takes crypto.
Together 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 | Together AI per hour | GPUBank per hour | Difference | 100 h at Together AI | 100 h at GPUBank |
|---|---|---|---|---|---|
 B200 SXM180 GB |
$8.19 | $4.15 | −49 % | $819.00 | $415.00 |
 H200 SXM141 GB |
$5.99 | $2.49 | −58 % | $599.00 | $249.00 |
 H100 SXM80 GB |
$3.99 | $1.85 | −54 % | $399.00 | $185.00 |
Together's cluster prices are per GPU in clusters of 8 GPUs or more, billed hourly, taxes excluded. Node vCPU, RAM and NVMe are not itemised. Its $3.99 dedicated H100 endpoint is a promotion until 30 Sep 2026.
A fine-tune, 40 h on an H100 SXM
Together AI
$159.60
GPUBank
$74.00
Saves $85.60 at GPUBank (−54 %)
A long-context job, 40 h on an H200
Together AI
$239.60
GPUBank
$99.60
Saves $140.00 at GPUBank (−58 %)
A full day, 24 h on a B200
Together AI
$196.56
GPUBank
$99.60
Saves $96.96 at GPUBank (−49 %)
02Their catalogue
Every GPU Together AI lists.
Together AI lists 8 GPU models. The 4 below are the ones recorded in our check, each at its lowest public price per GPU-hour. 3 of them are also rented at GPUBank (in blue).
- B300270 GB · on demand$9.99
- B200180 GB · on demand$8.19
- H200141 GB · on demand$5.99
- H100 SXM80 GB · on demand$3.99
At GPUBank
vCPU, RAM, NVMe and traffic included. All prices
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.
Together AI
2 h 14 m 03 s on H100 SXM
- GPU · $3.99 / h
- $8.91
- vCPU and RAM
- Not published
- Storage
- $0.16 per GiB-month, 1 TiB minimum
- Egress
- Included on clusters
- Public IPv4
- Not published
- Billing granularity
- Hourly on clusters, per minute on endpoints
- Minimum charge
- 8 GPUs per cluster; reservations paid upfront
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.
Together AI
Billed hourly until deleted, no pause.
One line of the billing page says on-demand clusters can be stopped and started.
GPUBank
Only the persistent disk, $0.03 per GB-month. Compute, IPv4 and scratch cost nothing.
04Where the GPUs run
5 countries against 4, and where they are.
Countries where Together 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.
United States
France
Netherlands
Sweden
Romania
Paris
Frankfurt
Ashburn
Singapore
Together AI5
- United States
France
Netherlands
Sweden
Romania
GPUBank4
- Paris
Frankfurt- Ashburn
Singapore
Together AI · 5 countriesGPUBank · 4 regions
05Paying and signing up
How each one takes your money.
Together AI
Payments and sign-up
Payment methods
ACH direct debit
No cryptoMinimum top-up $5 credit purchase
To start
- 1Google or GitHub OAuth account
- 2Credit or debit card
- 35 USD minimum credit purchase
Billing modelPrepaid credits
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.
| Together AI | GPUBank | |
|---|---|---|
| SOC 2 | Type II | Not published |
| ISO 27001 | Yes | Not published |
| HIPAA | Not published | No |
| GDPR DPA | DPA available | Published DPA, applies automatically |
| Availability target | No public uptime SLA | 99.9 % monthly, per region |
| Status page | status.together.ai | Public status page |
| Support | Tickets, email, Discord, shared Slack | 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 Together 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 opens with per-token model prices; GPU Clusters come after Provisioned Throughput and Dedicated Inference.
- 2The cluster table states all prices are per GPU per hour, with preemptible, on-demand and reserved columns up to 181+ days.
- 3The dedicated H100 carries a PROMO badge: $5.49 struck through, $3.99, valid until 09/30/26.
- 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
25 points, read off both sites.
GPUBank aheadTogether AI aheadEven
| Criterion | GPUBank | Together AI |
|---|---|---|
| Price and catalogue | ||
| H100 SXM, per GPU-hourReserved terms go down to $3.19; preemptible is $1.99. | $1.85, all-in | $3.99 (8-GPU cluster multiples) |
| Cheapest GPU-hourLowest on-demand price. Preemptible H100 is $1.99; GPUBank interruptible H100 is $1.11. | $0.23, RTX 4090 | $3.99, H100 SXM cluster |
| GPU models listedOnly H100, H200, B200 and B300 have a published price. | 10 | 8 |
| What the GPU price includesCluster prices are per GPU; node vCPU, RAM and local NVMe are not itemised. | vCPU, RAM, NVMe, IPv4, traffic | Not published |
| Billing | ||
| Billing granularity | Per second | Hourly on clusters, per minute on endpoints |
| Minimum chargeReservations are non-refundable and cannot be paused. | None | 8 GPUs per cluster; reservations paid upfront |
| Stopped instanceOne line of the billing page says on-demand clusters can be stopped and started. | Disk only | Billed hourly until deleted, no pause |
| Persistent storage | $0.03 per GB-month | $0.16 per GiB-month, 1 TiB minimum |
| Egress traffic | Included | Included on clusters |
| Public IPv4 | Included | Not published |
| Billing model | Prepaid balance | Prepaid credits |
| Payment and sign-up | ||
| Crypto paymentsStripe is the listed payment processor. | Bitcoin, USDT, Ethereum and 12+ more | Not offered: cards and US ACH |
| Card paymentsPrepaid cards refused, virtual cards often blocked. | No | Visa, Mastercard, American ExpressAhead |
| Minimum deposit | $25 | $5 credit purchaseAhead |
| Platform | ||
| API and CLI | REST API, CLI, Python SDK | REST API, tg CLI, Python and TypeScript SDKs |
| Terraform | Official provider | Official provider |
| Serverless GPUs | No | Per-token inference and dedicated endpointsAhead |
| Managed Kubernetes | No | Managed, or Slurm on KubernetesAhead |
| Ready-made templatesTemplates and Jupyter are not documented for GPU clusters. | 25 | Not published |
| Interruptible or spotH100 preemptible is $1.99 there; GPUBank interruptible H100 is $1.11. | Yes, 40 % off | Preemptible, in public preview |
| GPUs per instance | Up to 8 | 8 to 256 per cluster, 4,000+ reservedAhead |
| Multi-node interconnect | Not offered | InfiniBand, speed not publishedAhead |
| Network and regions | ||
| RegionsThe region list comes from an authenticated API; docs show us-central-8 and us-central-2 as examples. | 4 | Not published |
| CountriesThe European countries are named as capacity locations, not as self-serve regions. | France, Germany, US, Singapore | US, France, Netherlands, Sweden, RomaniaAhead |
| EU regionsEU residency needs dedicated or VPC setups. | Paris, Frankfurt | No self-serve EU region listed |
| Trust and support | ||
| Availability targetUptime guarantees only through an order form. | 99.9 % monthly | No public uptime SLA |
| Compliance reports | None published | SOC 2 Type II, ISO 27001:2022Ahead |
| Public status pageIt lists no GPU cluster component. | Yes | status.together.ai |
| Support | Tickets from a $25 balance | Tickets, email, Discord, shared SlackAhead |
09The right fit
Different needs. A clear choice.
What Together AI does better
- Self-serve InfiniBand clusters of 8 to 256 GPUs, 4,000+ reserved
- B300 on demand at $9.99, GB200 and GB300 NVL72 through sales
- Serverless inference, dedicated endpoints and fine-tuning on one account
- SOC 2 Type II and ISO 27001:2022 listed on the Trust Center
- Managed Kubernetes or Slurm, cards and US ACH payment
Choose GPUBank if
- H100 SXM at $1.85 per GPU-hour, against $3.99 on a Together cluster
- Rent a single GPU; Together clusters start at 8 GPUs
- Per-second billing, and a stopped instance costs only its disk
- Pay in Bitcoin, USDT and 12+ other coins, sign up with email
- RTX 4090, RTX 5090, L40S, A100 and MI300X, which Together does not price
Moving from Together AI takes an afternoon.
Together charges no egress fees on GPU clusters, and its jobs are standard Kubernetes or Slurm workloads with your own container images. Shared volumes persist after a cluster is deleted, so data can be copied out at any time.
- 1Launch the same GPU on GPUBank with a persistent disk in the region you need.
- 2Copy the shared volume with kubectl cp or rsync, or push it to S3 from a pod and pull it with rclone.
- 3Run the same container image, then delete the Together cluster and its volume, which bills separately.
# 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. Together lists the H100 SXM at $3.99 per GPU-hour on demand, in clusters of 8 GPUs or more, and $3.19 to $3.69 on reserved terms. GPUBank charges $1.85 per GPU-hour for a single H100 SXM, with vCPU, RAM, NVMe, IPv4 and traffic included.
Not as a general-purpose machine. Together GPU clusters are created in multiples of 8 GPUs. Its dedicated inference endpoints run on one GPU, but they serve a model through its API and are not a VM. GPUBank rents 1 to 8 GPUs per instance.
No. The pricing page marks the $3.99 dedicated inference H100 as a promotion valid until 30 Sep 2026, next to a struck list price of $5.49. The $3.99 on-demand cluster price is a separate, regular price.
No. Together takes Visa, Mastercard and American Express cards, and ACH from US banks, with a $5 minimum credit purchase. GPUBank is paid from a USD balance topped up in Bitcoin, USDT, Ethereum and more than a dozen other coins, from $25.
Together AI. It creates InfiniBand clusters of 8 to 256 GPUs on demand, up to 4,000+ reserved, with managed Kubernetes or Slurm. GPUBank rents up to 8 GPUs per instance and has no managed Kubernetes.
Without. Every price on this page excludes VAT and sales tax, for Together 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 Together AI was read on its own pages or official documentation. Prices last checked on 29 September 2026.
- 01together.ai/pricingGPU cluster, dedicated inference and storage prices
- 02together.ai/gpu-clustersScale limits, VRAM, reserved terms, capacity countries, SOC 2 and ISO badges
- 03docs.together.ai/docs/gpu-clusters-overviewCapacity types, preemptible, storage tiers, InfiniBand, hardware
- 04docs.together.ai/docs/gpu-clusters-billingBilling, non-refundable reservations, no egress fees, pause policy, credit exhaustion
- 05docs.together.ai/docs/preemptible-computePreemption mechanics and metering
- 06docs.together.ai/docs/cluster-storagePersistent vs ephemeral storage
- 07docs.together.ai/docs/gpu-clusters-quickstartReserved 1-90 days, 1 TiB minimum volume, SSH access
- 08docs.together.ai/docs/dedicated-endpoints/pricingPer-minute replica billing, hardware price table
- 09docs.together.ai/docs/billing-creditsPrepaid model, 5 USD minimum, no free trial
- 10docs.together.ai/docs/billing-payment-methodsCards, ACH limits, card-on-file rule
- 11docs.together.ai/docs/billing-troubleshootingPrepaid/virtual cards refused, 0 USD authorization hold
- 12docs.together.ai/docs/account-managementGoogle/GitHub OAuth sign-up
- 13docs.together.ai/reference/clusters-createGPU type enum, multiples of 8, billing types, Terraform mention
- 14docs.together.ai/reference/cli/clustersCLI, beta status, example region
- 15docs.together.ai/docs/gpu-clusters-apiCLI, REST, SkyPilot
- 16github.com/togethercomputer/terraform-provider-togetherOfficial Terraform provider
- 17docs.together.ai/docs/dedicated-container-inferenceCustom Docker inference, access via sales
- 18docs.together.ai/docs/supportSupport channels
What we could not verify (9)
- Year founded not stated on the official pages read (the Jul 2026 Series C post says the company was started 'four years ago').
- Per-node vCPU, RAM and local NVMe for cluster GPU types not published on the pricing or docs pages read.
- Full region list and per-region GPU availability are only returned by the authenticated regions API; public docs show example regions us-central-8 and us-central-2. France, Netherlands, Sweden and Romania are named as European capacity locations, not as self-serve regions.
- Prices for RTX_6000_PCI, L40_PCIE and H100_SXM_INF cluster GPU types (listed in the API enum) not published; the exact RTX 6000 variant (Ada or PRO) is not stated.
- Reserved 181+ days, reserved B300, dedicated inference H200/B300/GB300, and GB200/GB300 NVL72 cluster prices: contact sales.
- IPv4 / public IP policy not published.
- HIPAA: the GPU clusters page mentions choosing regions for HIPAA requirements, but no HIPAA attestation or BAA is published.
- No public SLA; the terms say no uptime guarantee unless agreed in an order form.
- Snapshots, Jupyter and templates are not documented for GPU clusters.
In short
GPUBank vs Together 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
- Together AI: $3.99 (8-GPU cluster multiples)
GPUBank: $1.85, all-in - Cheapest GPU-hour
- Together AI: $3.99, H100 SXM cluster
GPUBank: $0.23, RTX 4090 - Billing granularity
- Together AI: Hourly on clusters, per minute on endpoints
GPUBank: Per second - Stopped instance
- Together AI: Billed hourly until deleted, no pause
GPUBank: Disk only - Persistent storage
- Together AI: $0.16 per GiB-month, 1 TiB minimum
GPUBank: $0.03 per GB-month - Egress traffic
- Together AI: Included on clusters
GPUBank: Included - Crypto payments
- Together AI: Not offered: cards and US ACH
GPUBank: Bitcoin, USDT, Ethereum and 12+ more - Card payments
- Together AI: Visa, Mastercard, American Express
GPUBank: No - Minimum deposit
- Together AI: $5 credit purchase
GPUBank: $25 - GPU models listed
- Together AI: 8
GPUBank: 10 - Regions
- Together AI: Not published
GPUBank: 4 - Serverless GPUs
- Together AI: Per-token inference and dedicated endpoints
GPUBank: No - Availability target
- Together AI: No public uptime SLA
GPUBank: 99.9 % monthly - Compliance reports
- Together AI: SOC 2 Type II, ISO 27001:2022
GPUBank: None published - Support
- Together AI: Tickets, email, Discord, shared Slack
GPUBank: Tickets from a $25 balance - Headquarters
- Together AI: San Francisco, United States
GPUBank: Tallinn, Estonia
How this page was made
Every price and policy of Together 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. Together AI, its logo and its website belong to their owner and are shown only to identify the service compared. GPUBank is not affiliated with Together AI.
