ComputePrices.com review
Updated September 28, 2026 · DeployCue Editorial Team · Facts checked September 26, 2026
We earn a commission when you sign up through links on this site. Partner status can affect where a brand appears in our rankings.
Our verdict
ComputePrices.com suits developers and small teams who want a wide GPU catalogue at the lowest starting rate, with GPU hour from $0.015 and a free tier API plan beside it. The trade-off is billing granularity of hourly, which suits steady runs that fill their hours rather than quick bursts that do not.
Key facts
- Best for
- Developers and small teams wanting a wide GPU shortlist at the lowest starting hourly rate.
- GPU models
- H100 SXM, A100 SXM, H200, L40S, RTX PRO 6000, B200, HGX B300, RTX 6000 Ada, RTX A6000, RTX 4090, A100 PCIE, L4, RTX 5090, RTX A4000, L40, GB200, GB300, MI300X, A10, GTX 1650 (Source, checked September 26, 2026)
- GPU hour from
- $0.015 (Source, checked September 26, 2026)
- Billing granularity
- Hourly (Source, checked September 26, 2026)
- Free credits
- 750 requests/day (Source, checked September 26, 2026)
- Legal entity
- Lansky Tech (Source, checked September 26, 2026)
- Support channels
- Email, Help-centre (Source, checked September 26, 2026)
- Pricing model
- Free tier
- Official website
- computeprices.com
LLM APIs
Apps and profiles
- Elsewhere
- LinkedIngithub.com
Pros and cons
Pros
- GPU hour from $0.015 is the lowest starting rate in the set of 95 brands compared, which is the clearest reason to look here.
- The GPU catalogue covers H100 SXM, A100 SXM and H200, a broader mix than PoolCompute's NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe or SberCloud's B300, H100 and A100.
- API usage runs on a free tier plan that includes 750 requests/day, so light traffic can start without spend.
- Billing granularity of hourly is the mainstream shape here, with hourly billing the common choice across providers at 38.5%.
Cons
- Billing granularity is hourly, so a run that stops partway through still counts the whole unit.
- Support runs on email and help centre only, so an incident waits on written replies.
- The API sits on the free tier plan with 750 requests/day, so a high-traffic endpoint passes the allowance and moves onto paid rates.
- SberCloud's list, B300, H100 and A100, includes older parts such as V100, and the catalogue here, H100 SXM, A100 SXM and H200, does not name them.
Who it suits
- Developers and small teams that compare hourly GPU cost before anything else, since the entry rate here is the lowest in the set.
- Teams prototyping an API integration whose traffic fits inside 750 requests/day.
- Shoppers who have already weighed PoolCompute's NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe and want a wider list of parts to choose from.
Who should look elsewhere
- Buyers who want a fixed hyperscaler catalogue instead of breadth; SberCloud's B300, H100 and A100 is the shorter, more focused list.
- Shoppers chasing the cheapest consumer cards; Salad's NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti reaches older consumer and AMD classes that this list leaves out.
Pricing
Price is where this brand separates itself from the ones next to it. GPU hour from $0.015 is the lowest starting rate in the group, and the same figure carries through to renewal: it reads as the standing rate for an hour of card time rather than a limited opening offer, so entry and renewal do not drift apart. Readers who budget by the hour can size a month of compute from that number alone. The API allowance is 750 requests/day, a recurring daily amount rather than a trial credit, which covers steady light traffic before anything is billed. Taken together, the two figures answer both halves of the opening question: what a run costs while it is running, and what it costs to keep an integration alive between runs.
Who runs it
ComputePrices.com is operated by Lansky Tech, so a named company stands behind the price list and can be checked before spend is approved. Support runs on email and help centre, which means answers arrive in writing rather than by voice. That shape fits developers and small teams that file a ticket with logs, a container spec and a step-by-step reproduction, because the exchange stays searchable and nothing is lost the way a short call would lose it. It is a thinner channel set for a production incident, where a stalled training run needs a quick answer and the reply comes back in a thread. Procurement gets a named entity to check, which is more useful than a bare label on an invoice.
How ComputePrices.com compares
ComputePrices.com vs BytePlus9.6 vs 8.7
ComputePrices.com vs RunPod9.6 vs 9.5
ComputePrices.com vs Zhipu AI (Z.ai)9.6 vs 9.4
ComputePrices.com vs WaveSpeedAI9.6 vs 8.8
ComputePrices.com vs AIS Cloud9.6 vs 8.1
Frequently asked questions
What does GPU time cost on ComputePrices.com?+
The published starting rate is GPU hour from $0.015, and it is the lowest starting rate in this comparison. That number reads as the standing rate for an hour of card time rather than a limited opening offer, so the rate later is unchanged. Datacenter parts such as H100 SXM or B200 sit above this entry point.
Is there a free tier for the API?+
The API runs on a free tier plan, and that plan includes 750 requests/day. The allowance is a recurring daily amount rather than a trial credit, so a service with steady light traffic can keep running on it, while a busy endpoint will pass into paid usage. It suits early integration work, where the priority is checking how the endpoint behaves before spend matters.
Which GPUs are on offer?+
The catalogue covers H100 SXM, A100 SXM and H200. Datacenter parts sit next to workstation and consumer cards, so a team can match a card to a workload rather than take what is available. For contrast, SberCloud lists B300, H100 and A100 and PoolCompute lists NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe, both shorter and more focused on a few families.
How is GPU usage billed?+
Billing granularity is hourly, so time is metered in hourly units. For a long run that is simple to project from the published GPU hour from $0.015, and the cost of a job is known before it starts. The trade-off is that a job which stops between units still consumes the unit it entered, which matters for quick experiments.
Who is behind ComputePrices.com?+
The service is operated by Lansky Tech, a named legal entity rather than an anonymous marketplace listing. That gives a procurement team something concrete to check before a contract, and it tells a developer where documentation and support requests land. The support side runs on email and help centre, under the same company name.
What help is available when a job fails?+
Support runs on email and help centre, so a failure is handled in writing rather than by voice. That suits debugging, because a ticket can carry logs, a container spec and a reproduction step that stay searchable later. The listed set is written channels, so an incident that needs fast back-and-forth takes longer to settle than a live conversation would.
How does it compare with SberCloud and PoolCompute?+
Breadth is the dividing line. This brand's catalogue covers H100 SXM, A100 SXM and H200, SberCloud lists B300, H100 and A100 and PoolCompute lists NVIDIA Tensor Core H100 SXM, NVIDIA Tensor Core H100 NVL and NVIDIA Tensor Core H100 PCIe. A team that needs a particular card can find it here sooner, while a team that only wants H100 variants in separate form factors may prefer a tighter list. Entry price still points at ComputePrices.com.
Sources
- computeprices.com/docs/api checked September 26, 2026
Backs: Free credits, Pricing model, Support channels
“Free keys get 750 requests/day”
- computeprices.com/gpu checked September 26, 2026
Backs: Billing granularity, GPU hour from, GPU models
“Compare hourly rates across 82+ providers”
- computeprices.com/llms.txt checked September 26, 2026
Backs: Legal entity
“The site is operated by Lansky Tech. It is not affiliated with any cloud provider.”
Compiled from 3 source pages, checked September 26, 2026. We did not open an account.
Not confirmed yet: Egress fee.
ComputePrices.com
Developers and small teams wanting a wide GPU shortlist at the lowest starting hourly rate.