OpenRouter
A single account and bill for calling hundreds of AI models from different providers, so switching models doesn't mean signing a new contract.
What you'd use it for
Building an app or agent that calls AI models through code, and wanting one API key and one invoice instead of a separate account for every model provider.
- Call GPT, Claude, Gemini and hundreds of other models through one API key and one request format
- Switch which model a feature uses by changing a model name string, without a new integration
- Test a model for free, up to 50 requests a day, before committing spend to it
- See published usage rankings showing which models other developers are actually calling in production
- Pay with credit card, AliPay or crypto and get one consolidated bill instead of separate provider invoices
The verdict
OpenRouter is the right call for anyone shipping code against more than one model: one key, one invoice, and published usage rankings that show what other developers are actually running in production. The catch is the fee itself — 5.5% to 8% stacked on top of whatever the underlying provider already charges — so it is never the cheapest way to call a single model you have already committed to. It suits developers and small teams who want the freedom to swap models without renegotiating a contract, more than anyone chasing the lowest per-token price.
What it costs
| Tier | Price | For | Includes |
|---|---|---|---|
| Free | $0 platform fee | Trying models out before committing to a paid workload. | 50 requests/day across 25+ free models and 4 free providers. |
| Standard | 5.5% fee on top of each model's underlying cost, pay-as-you-go | A business calling models directly from its own app or agent. | Access to 500+ models across 80+ providers, no monthly minimum, no annual commitment. |
| Business | 8% fee on top of each model's underlying cost, pay-as-you-go | Teams that need regional routing and stronger delivery guarantees. | Everything in Standard, plus regional routing and enhanced controls. |
| Enterprise | Custom pricing, negotiated fee discounts | Organisations with high, committed usage volume. | Dedicated support, invoicing, negotiated fee reductions. |
Read from OpenRouter's own pricing page on 25 Sep 2026. Vendors change prices without notice.
What you need to start
- An OpenRouter account and an API key created in account settings — no sales contact needed
- Your own app, agent or script that can send HTTP requests, since there is no built-in chat interface
- A credit card, AliPay or crypto payment once you go past the free tier's 50 requests a day
Minutes to a working API key and a free-tier test request, since sign-up needs no credit card and no sales call; adding paid credits and wiring it into your own app is the same sitting for most people.
Your first run Following the quickstart, you get an API key from your account settings, send one POST request naming a model such as openai/gpt-4o, and get back a normal chat completion within minutes — no separate account or key for the underlying model provider.
Where it breaks
The percentage fee sits on top of whatever the underlying model provider already charges, so there is no flat number to quote until you know which models and how much volume you're running.
A browser dashboard for keys, billing and usage — the actual work happens through your own app or agent calling OpenRouter's API, not through a chat interface of its own.
First entry in this reference. There is no previous check to compare it against.
When not to use it
Anyone who just wants to talk to an AI in a browser — that job belongs to ChatGPT or Claude's own app, not an API router. Also not for a team that has already settled on one model provider and has no plans to switch, since the fee is pure overhead with no optionality to show for it.
OpenRouter is not a tool you sit inside. It is the account behind the tool — a router that lets one API key and one invoice reach hundreds of models from dozens of providers, so an app built on ChatGPT today can call Claude tomorrow without a new contract, a new billing relationship, or new code beyond a model name.
That framing matters for who this is for. If you talk to an AI directly in a browser tab, OpenRouter has nothing to offer you — ChatGPT or Claude’s own apps do that better. OpenRouter earns its place when something you’ve built, or had built, needs to call a model programmatically and you want the freedom to change which model that is without renegotiating anything.
What the fee actually is
There is no flat monthly price. OpenRouter passes through each provider’s own per-token cost and adds a platform fee on top — 5.5% on the Standard plan, 8% on Business, with Enterprise customers able to negotiate it down at volume. The free tier exists to test the plumbing, not to run real work: 50 requests a day across a rotating set of free models.
The honest trade-off
Routing through OpenRouter costs more per token than going straight to a provider, because of the fee. What it buys back is optionality — one integration instead of five, published usage rankings across the industry that make it easy to see which models people actually rely on, and the ability to swap a failing or discontinued model for another without touching your app’s billing setup.
Alternatives
Assistants & chatModelPoe
One subscription and one inbox for chatting with ChatGPT, Claude, Gemini and dozens of other AI models, so you stop paying for and switching between separate apps.
Assistants & chatAppsHuggingChat
A free chat interface over open-weight AI models, with no subscription required, useful for testing a model before deciding whether to pay anyone for it.
Assistants & chatAppsAbacus.AI ChatLLM
One subscription that bundles access to most frontier AI models plus document, image and small agent tools, instead of paying separately for ChatGPT, Claude and others.
Get OpenRouter
Buying the tool is not the hard part.
The ebook is about what to hand over, in what order, and how you find out when it goes wrong. One payment, no subscription.