Terms of Service

1. What model.cards is

model.cards is an API gateway and router. You connect API keys from your own provider accounts, send us a chat-completion request naming a model (or modelcards/auto, which picks one per request by complexity), and we forward it to an upstream provider using your key. The provider bills you directly at your own rates; we record token counts and a cost estimate for your analytics. We do not host or train models ourselves, and we do not resell inference.

The models, their prices and their behaviour belong to the upstream providers. When you call one through us, that provider's own terms and content policies apply to your request in addition to ours.

2. Beta status

The service is in beta. Concretely, that means:

• There is no uptime commitment and no SLA. We may take the gateway down for maintenance without notice.
• Features can change or be withdrawn. Anything labelled “coming soon” is not a commitment to ship.
• Rate limits are fair-use rather than published quotas. We will contact you before throttling an account doing normal application traffic.
• Some figures on the site — provider uptime, latency, throughput, and the token volumes behind the rankings — are reference data used to populate the interface. They are labelled where they appear and are not measurements of traffic we have served. Do not rely on them for procurement decisions.

3. Your account

You need an account to get an API key. Keep your credentials and your API keys secret: any request carrying one of your model.cards keys is treated as yours and runs on the provider keys you connected. If a key leaks, disable or delete it from the dashboard — that takes effect immediately.

You must be old enough to enter a contract where you live, and you may not use the service if we have previously terminated your account.

4. Your provider keys and money

• You pay your providers, not us. Requests run on the provider API keys you connect, so the provider invoices you directly at your account's own rates. model.cards holds no balance, takes no fee per token, and is free during beta.
• Stored keys are encrypted. Connected provider keys are encrypted at rest, decrypted only to authenticate your own requests, and deleted when you remove them or close your account.
• Cost figures are estimates. The costs shown on responses and in your usage log are computed from provider-reported token counts and catalog prices. Your provider's invoice is the authoritative bill; if the two disagree, the invoice wins.
• Spend guardrails are best-effort protection, not a billing guarantee. They stop requests whose projected cost would pass a limit you set, but your agreement with your provider governs what you actually owe them.
• Routing decisions are ours to make well. modelcards/auto chooses among the providers you connected. You can always pin a specific model instead.

5. Acceptable use

What you may and may not send through the gateway is set out in the Acceptable Use Policy, which forms part of these terms. Upstream provider policies apply on top of it. We can suspend a key or an account that breaks either, and we will tell you why when we do.

6. Your content

You keep every right you have in what you send us and in what the models send back. We claim no ownership. We use your requests only to serve them, to bill them, and to keep the service running — never to train models. The Privacy Policy describes exactly what we store and for how long.

You are responsible for what you send and for what you do with the output. Model output can be wrong, biased or fabricated; check it before it reaches a decision that matters.

7. What we do not promise

The service is provided as-is. We do not warrant that it will be uninterrupted, that a particular model or provider will stay available, that prices will not change when a provider changes theirs, or that output will be accurate or fit for your purpose.

8. Liability, in plain language

If we get something wrong and it costs you money, our total liability to you is capped at whatever you paid us in the three months before the problem — during the free beta, that is zero, so choose what you route through us accordingly. We are not liable for lost profits, lost data, or knock-on business losses, even if you told us they were possible.

Nothing here limits liability that cannot be limited by law, such as for fraud or for death or personal injury caused by negligence.

9. Suspension and termination

You can close your account at any time from the dashboard. We can suspend or close an account that breaks these terms, that puts the service or an upstream provider at risk, or that we are required to act on by law. Closing your account deletes your stored provider keys with it.

10. Changes

We may update these terms as the product changes. The date at the top of this page always reflects the current version, and we will give notice by email before a change that materially reduces your rights takes effect. Continuing to use the service after that means you accept it.

11. Governing law

These terms are governed by the laws of [Singapore], and the courts of [Singapore] have exclusive jurisdiction over any dispute arising from them.

12. Contact

Questions about these terms: support@model.cards. Contracts, invoicing and design partnerships: sales@model.cards.