Legal

The documents

Four pages, each answering one question a gateway should be able to answer plainly.

PrivacyRequest metadata is recorded; prompts and replies are not. The one exception, named. TermsWhat free capacity really is, and the promises we deliberately do not make. Sub-processorsHosting, model providers, authentication and email: who receives what. SecurityCredential handling and transport, and the list of what this service does not have.

The one-line answer to the usual question

Can FreeModel read my prompts? They are not stored, so there is nothing to read after the request finishes. The request itself is forwarded to the provider serving the model, which processes it under its own terms. The single place where request content is written to a database is the agent endpoint, described in full on the privacy page.

Privacy: what is recorded

One question, one answer. What is recorded when a request passes through the gateway is metadata (model, token counts, HTTP status, duration), plus a single named exception where request content is written to a database. Prompts and replies are not stored.

Terms: what is promised

The terms are short because the product is short: you send a request, we route it, a provider answers. What they deliberately do not contain is a service level, an uptime figure, or a promise that any particular model stays available next month.

Sub-processors: who is involved

A gateway is a chain of third parties. This document names them — the hosting providers, the model providers and routing services that receive your prompt, and the authentication and email providers behind the account system — and says what each one receives.

Together the three documents describe a gateway that reaches 200+ models: 153 chat models, 77% of the list, and the remaining 23% split across image, video, speech synthesis, speech recognition, embeddings and rerank.[1] Six of the 199 are modality routes rather than single models (3% of the list) and a route resolves to a provider at request time rather than naming one in advance.[1] That property is the reason the sub-processor list is written by category instead of being frozen: the set of parties that can receive a given prompt changes as supply changes.

What these documents do not cover

They govern the gateway, and they stop at its edges. What a provider does with your prompt while it processes the request is covered by that provider's own terms, not these. A third-party client you point at the gateway (an editor extension, a CLI, an agent framework) keeps whatever local history it keeps, and that is outside our control too.

Legal FAQ

Which of the four should I read first?

Privacy, if your question is "where does my prompt go". Terms, if your question is "what happens when a request fails". Sub-processors, if you have to name the parties involved for a review or a procurement form. Security, if the form asks what certifications exist — that page answers it with a "no" rather than a paragraph.

What do these pages deliberately not promise?

No SLA, no uptime figure, and no retention period we would not actually enforce. Each is stated plainly on the page it belongs to rather than left vague. A gateway that publishes a number it cannot measure is worse than one that says nothing at all.

Do the documents cover the console, or only the API?

Both. Privacy applies to the whole service, including account data such as your email address. Terms apply to the gateway and the console together: an account is needed to obtain a key, and the key is bound by the same rules wherever it is used.

Is FreeModel allowed to read what I send?

There is nothing stored to read after the request finishes, because prompts and replies are not written to the database. The request itself is forwarded to the provider serving the model, which processes it under its own terms. The single exception is the agent endpoint, described in full on the privacy page.

Where do I raise something these pages do not answer?

postmaster@gladec.cn for anything about data, and the same address for everything else. Grievances first, please — most things turn out to be a bug rather than a dispute.

How do I know when one of them changes?

Each page carries a date at the bottom, and the date moves when the content does. A change that affects what is stored is also noted on the about page.

FreeModel is a product of Aiglade. Last updated 2026-09-21.

Where the percentages come from

Every share quoted on this site is counted from the live model list, not estimated. Divide the two counts in the last column, round to the nearest whole percent, and you get the number in the first.

Share	Of what	Count
77%	are chat models	153 of 199 listed entries
9%	generate images	18 of 199 listed entries
23%	are not chat models	46 of 199 listed entries
3%	are modality routes	6 of 199 listed entries
17%	carry a context length	34 of 199 listed entries

[1] Counted on 29 September 2026 from GET https://freemodel.online/v1/models, which needs no key and no account. Every row uses all 199 listed entries as the denominator, because the only thing the endpoint states for every entry is its modality. It no longer publishes per-model capability flags such as tool calling or reasoning, so this site quotes no shares for them. Fetch the list and recount whenever you want to check the arithmetic.