llmfa.st
HomePrivacy

Terms of Use

Last updated: 21 August 2026

1. This agreement 2. The Services 3. Third-party models 4. Accounts and API keys 5. Fees and payment 6. Late payment 7. Refunds 8. Cancellation 9. Your content and outputs 10. Output accuracy 11. Content restrictions 12. Prohibited activities 13. Rate limits and capacity 14. Security 15. Feedback 16. Intellectual property 17. Third-party platforms 18. Privacy 19. Term and termination 20. Changes 21. Disclaimers 22. Limitation of liability 23. Indemnification 24. Trade controls 25. Governing law 26. Notices 27. Miscellaneous 28. Contact

1. This agreement

These Terms of Use (the "Terms") are a binding agreement between you — whether personally or on behalf of an organisation ("you", "your", "Customer") — and llmfa.st ("llmfa.st", "we", "us", "our").

They govern your access to and use of our inference API at api.llmfa.st, our website at llmfa.st, and any related documentation, tools or interfaces (together, the "Platform"). The Platform provides hosted inference on open-weight language models (the "Services"). Any additional policy we publish and reference here forms part of these Terms.

We do not represent that the Platform is appropriate or available in every country. If you access it from somewhere that would make doing so unlawful, or would subject us to registration requirements we have not met, you do so on your own initiative and are responsible for compliance with local law.

The Platform is not built to satisfy sector-specific regimes such as HIPAA, PCI-DSS or FedRAMP. Do not send us data governed by those regimes unless we have agreed otherwise in a signed written agreement.

You must be at least 18 and not a minor in your place of residence to use the Platform.

By creating an account, using an API key, or otherwise accessing the Services, you agree to these Terms. If you are accepting on behalf of an organisation, you confirm you have authority to bind it. If you do not agree, do not use the Platform.

2. The Services

Subject to these Terms and to your payment of applicable fees, we grant you a non-exclusive, non-transferable right to access and use the Services for the period you have paid for. That includes using our API to build the Services into your own applications, products or services (each a "Customer Application") and making those available to your own end users.

You are responsible for your end users' use of the Services through your Customer Application, and must ensure they are bound by terms at least as protective as these.

3. Third-party models

The Services run open-weight models published by third parties — for example the Qwen, DeepSeek, GLM and Kimi families. We do not own these models. Each is released under its own licence, and some carry restrictions on acceptable use, on training other models from their outputs, or on commercial deployment above a certain scale.

We identify the upstream source of every model we serve in the hugging_face_id field of our models endpoint, so you can review the applicable licence before you use it. You are responsible for complying with the licence of each model you call, and for ensuring your end users do the same.

The models are provided as-is. We did not train them, we cannot audit their training data, and we make no representation about their behaviour, accuracy or the provenance of their outputs.

4. Accounts and API keys

You represent and warrant that:

  • the registration information you give us is true, accurate and current, and you will keep it so;
  • you will keep your API keys confidential;
  • you are not a minor in your place of residence; and
  • your use of the Services will not breach any applicable law.

You are responsible for all activity carried out with your API keys, including activity you did not authorise but which resulted from your failure to keep them secret. Tell us immediately at [email protected] if a key is exposed and we will revoke it. We store only a cryptographic hash of each key and cannot recover a lost one — we can only issue a replacement.

Do not share an account or key with anyone outside your organisation, create multiple accounts to circumvent limits or free allowances, or register using disposable email addresses.

5. Fees and payment

Pricing for each model is published on the Platform and, authoritatively, at our models endpoint. Usage is billed on measured tokens: prompt tokens, completion tokens, and cached prompt tokens each at their published rate. Token counts are measured by the inference engine that served the request, and that measurement is authoritative for billing.

You authorise us to charge your chosen payment method for all amounts due. Where you have enabled automatic top-up, you consent to recurring charges when your balance falls below the threshold you set, until you disable it.

All amounts are in USD and are exclusive of VAT, sales tax and other duties, which you are responsible for unless you supply a valid exemption certificate. We may correct pricing errors even after a payment has been taken, and may change prices on 30 days' notice; reductions take effect immediately.

Charges you do not dispute in writing within 60 days of the charge date are treated as accepted.

Where you reach us through a routing platform such as OpenRouter rather than directly, billing is governed by our agreement with that platform and by your agreement with them, not by this section.

6. Late payment

An account is delinquent if an amount remains unpaid at the start of the next billing cycle. We may suspend Services on accounts delinquent for more than 15 days, and may archive or delete data on accounts delinquent for more than 30 days. Overdue amounts accrue interest at 1% per month or the maximum the law allows, whichever is lower, and you are responsible for reasonable costs of collection.

7. Refunds

Usage-based charges are for compute already consumed and are non-refundable. Unused prepaid credit is refundable on request within 12 months of purchase, less any amounts owed.

Nothing here limits any non-waivable statutory right of withdrawal or refund you may have as a consumer under the law of your country of residence.

8. Cancellation

You may stop using the Services at any time and close your account from your account settings or by contacting us. Where you hold a subscription, it continues to the end of the current term and then ends; if you do not cancel before the renewal date it renews for a further term of the same length.

9. Your content and outputs

You and your end users submit prompts and other material to the Services ("Input") and receive generated material back ("Output"). Input and Output together are your "Customer Content".

You keep everything. As between you and us, you retain all rights you have in your Input, and to the extent we hold any rights in Output generated for you, we assign them to you — subject always to the licence of the underlying model that produced them.

You grant us only the narrow, non-exclusive, royalty-free licence needed to receive, process and transmit your Input in order to serve your request and return the Output. That licence ends when the request ends.

We do not train models on your Customer Content. We do not sell it, share it for anyone else's commercial purposes, or use it to improve any model. There is no opt-out to configure, because there is nothing to opt out of.

We operate zero data retention for inference content. Prompts and completions are held in memory for the lifetime of a request and are never written to disk, to logs or to any database. What we do keep is request metadata — timestamp, model, token counts, latency and HTTP status — for 30 days, for billing and reliability monitoring. Our Privacy Policy sets this out in full.

You are responsible for your Input and confirm you hold all rights, consents and permissions needed to send it to us and to have it processed.

10. Output accuracy

Language models are probabilistic. Output may be wrong, incomplete, outdated, biased, or offensive, and may confidently state things that are untrue — including about real people, places and events. Output does not represent our views.

Because generation is probabilistic, Output is not guaranteed to be unique. Another customer sending a similar prompt may receive similar or identical content, and content generated for someone else is not yours regardless of how similar it is.

You must evaluate Output for accuracy and suitability before relying on it, including human review where the stakes justify it. Do not use Output as a sole source of truth or as a substitute for professional advice.

You must not use Output to make, or materially inform, decisions about an individual's credit, employment, education, housing, insurance, immigration status, legal position or medical care, and you must not use the Services in safety-critical systems where a failure could cause injury or death.

11. Content restrictions

You must not use the Services to generate, transmit or store material that:

  • sexualises minors in any way, or constitutes child sexual abuse material;
  • incites, threatens or facilitates violence, terrorism or genocide;
  • harasses, defames, intimidates or abuses any person;
  • assists in developing chemical, biological, radiological, nuclear or cyber weapons;
  • infringes anyone's copyright, patent, trademark, trade secret, privacy or publicity rights;
  • uses the name, voice or likeness of an identifiable person without their consent;
  • is designed to deceive at scale — disinformation campaigns, impersonation, fraud, phishing or spam;
  • discriminates against people on the basis of race, national origin, religion, gender, sexual orientation, disability or another protected characteristic; or
  • breaches the licence of the model that generated it.

You must not represent Output as human-authored where doing so would mislead someone to their detriment or breach a disclosure obligation that applies to you.

We may take steps we reasonably believe necessary to verify compliance with these Terms. We may preserve and disclose account information and, where we hold it, Customer Content, where we believe in good faith that doing so is reasonably necessary to comply with legal process, enforce these Terms, investigate fraud or a security incident, or protect the safety of any person.

12. Prohibited activities

You must not:

  • reverse engineer, decompile or attempt to extract the weights, architecture or configuration of any model we serve;
  • use Output to train, fine-tune or distil a competing model, where the underlying model's licence prohibits it;
  • resell raw API access, or operate the Services as a service bureau, without a written reseller agreement with us;
  • circumvent rate limits, quotas, billing or access controls, including by spreading one key across parties or creating multiple accounts;
  • probe, scan or test the vulnerability of the Platform, or breach any security or authentication measure, without our prior written consent;
  • interfere with the Platform or place an unreasonable load on it;
  • use automated systems to scrape the Platform beyond documented API use;
  • impersonate another person, or misrepresent your affiliation with any person or organisation;
  • harass or threaten our staff; or
  • remove any proprietary notice from anything we provide.

We may suspend access immediately, without notice, where continued use risks harm to a person, to us, or to the stability of the Services. Where circumstances allow, we will give notice and an opportunity to fix the problem first.

13. Rate limits and capacity

We apply per-key and per-model limits, published in the capacity fields of our models endpoint. When we are at capacity we return HTTP 429 with a Retry-After header immediately rather than holding your request in a queue.

We target 99.5% monthly availability but do not guarantee uninterrupted service under these Terms. Any binding availability commitment, and any service credits attached to it, must be set out in a separately signed service level agreement.

We may add, change, retire or replace models. Where we retire a model we will publish a deprecation_date in our models endpoint at least 30 days beforehand, except where a licence change, security issue or upstream withdrawal requires us to act faster.

14. Security

We maintain an information security programme with technical and organisational measures appropriate to the risk, including encryption in transit, hashed storage of credentials, network isolation of inference hardware, least-privilege access control and logging of administrative action. No system is perfectly secure, and we do not warrant that the Platform will be free from compromise.

15. Feedback

If you send us suggestions, ideas or other feedback about the Services, we may use it for any purpose without obligation or compensation to you. Do not send us feedback you consider confidential or proprietary.

16. Intellectual property

Customer Content is yours, as set out in section 9. Everything else on the Platform — our software, APIs, documentation, designs, text, graphics, and the llmfa.st name and logo — belongs to us or our licensors and is protected by intellectual property law.

We grant you a limited, revocable licence to access and use the Platform and our documentation for the purpose of using the Services. You may not copy, modify, distribute, sell, sublicense or create derivative works of our software, or use our name or logo in a way likely to cause confusion, without our prior written permission. All rights not expressly granted are reserved.

17. Third-party platforms

You may reach the Services through a third-party routing platform such as OpenRouter. Where you do, your relationship with that platform is governed by their terms, and their billing, data handling and support arrangements apply to that relationship. These Terms continue to govern our provision of inference and your use of the models.

Links from the Platform to third-party sites, models or documentation are provided for convenience. We do not control or endorse them, and your use of them is at your own risk and subject to their own terms.

18. Privacy

Our Privacy Policy explains what we collect, why, and how long we keep it. By using the Services you agree to it. If you use the Services to process personal data, you are the controller of that data: you must have a lawful basis, give the required notices, and honour the rights of the people concerned. Where we act as your processor, the terms of a data processing agreement apply — contact us to put one in place.

19. Term and termination

These Terms apply for as long as you use the Services. Either of us may terminate for convenience on 30 days' written notice. We may terminate or suspend immediately for material breach, for non-payment after notice, or where required by law or by a model licensor.

On termination your access ends, and any unused prepaid balance is refunded on a pro-rata basis less amounts you owe. Sections 3, 9, 10, 16, 21, 22, 23 and 25 survive termination, along with any other provision that by its nature should.

If we terminate your account for breach, you may not open another without our written consent.

20. Changes

We may update these Terms. We will post the updated version with a new date and, for changes that materially affect your rights, give at least 30 days' notice by email to active customers before they take effect. Continuing to use the Services after that date means you accept the change. If you do not accept it, stop using the Services and close your account before it takes effect.

We may also modify or discontinue the Services, in whole or in part. We will give reasonable notice of a change that materially reduces functionality you are paying for.

21. Disclaimers

The Platform and Services are provided "as is" and "as available". To the fullest extent permitted by law we disclaim all warranties, express or implied, including implied warranties of merchantability, fitness for a particular purpose, title and non-infringement. We do not warrant that the Services will be uninterrupted, secure or error-free, that defects will be corrected, or that Output will be accurate, current, complete or fit for any purpose.

Nothing in these Terms excludes or limits liability that cannot lawfully be excluded or limited, including liability for death or personal injury caused by negligence, for fraud or fraudulent misrepresentation, or for any non-waivable statutory consumer right.

22. Limitation of liability

Subject to section 21, neither party is liable for indirect, incidental, special, consequential or punitive damages, or for lost profits, lost revenue, lost data or loss of goodwill, however caused and on any theory of liability, even if advised of the possibility.

Our total aggregate liability arising out of or relating to these Terms is limited to the greater of the fees you paid us in the 3 months immediately before the event giving rise to the claim, and USD 100.

Some jurisdictions do not allow these exclusions or limitations, in which case the ones that are not permitted do not apply to you and you may have additional rights.

23. Indemnification

You will defend, indemnify and hold harmless llmfa.st, its affiliates, and their officers, employees and agents against any third-party claim, and any resulting loss, damage, liability, cost and reasonable legal fees, arising out of your Customer Content, your use of the Services, your breach of these Terms, or your breach of any law or third-party right. We will notify you promptly of any such claim and give you reasonable cooperation in defending it.

24. Trade controls

You must comply with all applicable export control and sanctions laws. You may not use the Services, or allow them to be used, in or for the benefit of any embargoed country or territory, or by any person or entity subject to trade restrictions. You confirm that you are not such a person and are not acting on behalf of one.

25. Governing law and disputes

Before either of us starts proceedings, we will each try in good faith to resolve the dispute directly. Write to [email protected] setting out the problem and what you would like done about it, and we will respond within 30 days.

These Terms, and any dispute arising out of or in connection with them or the Services, are governed by the law of the jurisdiction in which llmfa.st is established, and its courts have jurisdiction — except that either party may seek injunctive relief in any court of competent jurisdiction to protect its intellectual property or confidential information.

Nothing in this section deprives you of the protection of any mandatory consumer law, or of the right to bring proceedings in the courts of your own country, where the law where you live gives you that right. The United Nations Convention on Contracts for the International Sale of Goods does not apply. No claim relating to the Services may be brought more than 2 years after the cause of action arose.

26. Notices

Notices to us must be sent by email to [email protected]. Notices to you will be sent to the email address on your account and are treated as received 24 hours after sending, unless we are told delivery failed. Keep your contact details current.

27. Miscellaneous

These Terms, together with the Privacy Policy and any signed order form or service level agreement, are the entire agreement between us about the Services. Our failure to enforce a provision is not a waiver of it. If a provision is held unenforceable, it is severed and the rest continues in force. You may not assign these Terms without our written consent; we may assign them in connection with a merger or a sale of substantially all our assets. Neither party is liable for delay or failure caused by events beyond its reasonable control. Headings are for convenience only.

28. Contact

llmfa.st
General enquiries: [email protected]
Legal notices: [email protected]
Security reports: [email protected]
Privacy requests: [email protected]

© 2026 llmfa.st
Home Privacy Policy Contact