⌕ search /
Terms of use

Terms of use

Estimates, not quotes. Results are model estimates; validate them with your own load test.

These terms govern your use of howmanygpus.ai (the "service"). howmanygpus.ai is operated by Siarhei Harlinski ("we", "us"). By using the service you agree to them. If you use it on behalf of a company, you confirm you may accept these terms for it.

1. The service

howmanygpus.ai is a free tool that estimates GPU memory, speed and cost for large language model inference, fine-tuning and edge deployment, and shows GPU rental prices collected from third-party providers. No account is required. We may add, change or remove features at any time.

2. Estimates, not quotes

All results — VRAM, throughput, latency, GPU counts, replicas, cost per hour, per 1M tokens or per session — are estimates produced by a model from published specifications and public measurements. They are not measurements of your workload, not a vendor quote and not a guarantee of performance. Actual results depend on the model, runtime version, drivers, configuration and traffic. Typical error is stated on the Accuracy page. Validate any configuration with your own load test before purchase, deployment or commitment. You are solely responsible for decisions you make based on the results, including purchases, rentals and deployments.

3. Prices are indicative and come from third parties

Prices are on-demand list prices collected from providers' public price lists and pricing APIs, normally once a day, and converted to US dollars where needed. They may be outdated, incomplete, region-specific or wrong, and they exclude discounts, reservations, taxes, egress, storage and support. Providers set their own prices and terms; we do not sell or broker compute. Always confirm a price with the provider before budgeting.

4. Acceptable use

You may use the service and its results for personal and internal business purposes, including quoting results in proposals, reports and articles, with a reference to howmanygpus.ai. You may embed the result card and subscribe to the RSS feeds.

You may not:

  • access the service with automated tools beyond the request rates it enforces (requests above the limit receive HTTP 429), or try to bypass those limits, for example by rotating addresses;
  • systematically download the catalog or the price history to build or feed another dataset, product or service;
  • resell, sublicense or redistribute the price data or the feeds, in whole or in substantial part, as a data product;
  • interfere with the service, probe it for vulnerabilities without permission, or use it to break the law.

We may block traffic that breaks these rules.

5. Our content

The site's text, design, catalog curation, models and collected price history are ours or our licensors'. Using the service gives you no ownership of them beyond the use allowed in section 4. Third-party names, logos and specifications belong to their owners; their appearance does not imply endorsement. howmanygpus.ai is not affiliated with NVIDIA, AMD or any provider shown.

6. Links and third-party services

The service links to provider sites, Hugging Face and other third parties. We are not responsible for their content, availability or terms.

7. No warranty

The service is provided "as is" and "as available", without warranties of any kind, express or implied, including accuracy, completeness, fitness for a particular purpose and non-infringement. We do not warrant that it will be uninterrupted or error-free.

8. Limitation of liability

To the maximum extent permitted by law, we are not liable for any indirect, incidental, consequential or special damages, or for lost profits, revenue, data or business opportunities, arising from your use of, or reliance on, the service or its results — including purchase, rental or capacity decisions. The service is free, so our total liability for any claim relating to it is limited to the amount you paid us for it — zero. Some jurisdictions do not allow these limits; in that case they apply to the extent allowed.

9. Changes to these terms

We may update these terms. Material changes will be noted on this page. Continuing to use the service after a change means you accept the updated terms.

10. Governing law

These terms are governed by the laws of Poland, without regard to conflict-of-law rules. Disputes go to the courts of Warsaw, Poland, unless mandatory consumer law gives you another forum.

11. Contact

Questions about these terms: Siarhei Harlinski — [email protected] · LinkedIn · X @howmanygpus.