One API key. Free models first. Automatic fallback.

What it is

• One key, many providers. You call one OpenAI- or Anthropic-compatible base URL instead of creating an account and a key at every model provider.
• Free-first routing. On every request, free-eligible models are tried first. Paid capacity sits behind them as the fallback.
• Automatic failover. If a provider returns a quota error, a rate limit or a hard failure, the gateway tries the next candidate instead of returning the error to your app.
• A public model list. The catalogue is readable at /v1/models without a key, so you can verify what is callable before you sign up.
• Per-token billing on paid calls. Free-eligible calls are not billed; paid calls are priced at the answering model's published rate. See Pricing.

What it is not

• Not a model provider. FreeModel does not train or host the models; it routes to the providers that do.
• Not a directory. A model list is evidence, not the product. The product is the callable API with free-first routing and failover.
• Not a chat app. It is an API gateway for your own clients, scripts, coding agents and applications.

Try it without a key

1. Call GET https://freemodel.online/v1/models. It is public; no key or account is needed to see the current model list and which rows are free.
2. Create an account with an email — no card is required for the free tier.
3. Read the free allowance and reset window in the console; it is shown per account and per tier.
4. Send your first request to the OpenAI- or Anthropic-compatible base URL with your gateway key.

The free tier is 800 rounds per month and 30 per day, with no card required. When the allowance runs out, requests fall through to paid capacity rather than failing outright.

Pricing at a glance

• Free: 800 rounds per month, 30 per day, no card required.
• Standard: $49 per month — flagship models and higher concurrency.
• Pro: $249 per month — pro-tier models and the highest limits.

Paid calls are billed per token at the answering model's rate. The full per-model table, free-eligible rows and billing rules are on the Pricing page. Check the current number there before committing; the page states its last-updated date.

How routing works

1. The request arrives at the OpenAI- or Anthropic-compatible base URL with your gateway key.
2. The gateway resolves the model name to a candidate list and puts free-eligible candidates first.
3. It sends the request to the first candidate. If that provider returns a quota error, a rate limit or a hard failure, the next candidate is tried.
4. If a free candidate answers, nothing is billed. If a paid candidate answers, the input and output tokens are priced at that model's rate.
5. Usage is written to the console, visible per key and per day.

You do not tag individual calls or choose a fallback list per request. Routing is automatic; see the billing walkthrough for the detail.

Who it is for

• Developers and small teams who want one endpoint and one key instead of managing many provider accounts.
• Coding-agent and CLI users who want a model to try free capacity first and fail over without changing the client.
• Prototypes and bursty workloads where free capacity is enough most of the time and paid capacity should only absorb the overflow.

If you need a guaranteed enterprise SLA, per-call manual routing, or a provider-specific feature that the gateway does not expose, use the provider directly. The gateway is for the common path, not every edge case.

Product FAQ

Start with the public list

Open /v1/models to see what is callable today, then create an account and make your first call on the free tier. No card, no key needed to inspect the list.

Open the console →