Pricing depends on the model you use. You can get a quick overview with prices in their documentation: https://openrouter.ai/docs#models

Or browse: https://openrouter.ai/models for a few more details.

As far as usage goes, the documentation I linked goes into it more deeply. But to quickly get started, and assuming you are familiar with the openAI API, you basically can use openAI api calls by changing the baseurl. https://openrouter.ai/docs#quick-start

edit:

I don't remember if this was turned on or off by default. But in your settings (https://openrouter.ai/account#default-model) there is an analytics setting:

> Allow logging of inputs & outputs and get a 1% discount on all LLMs

I have turned this off. Having said that, they effectively route the calls to the LLM providers and these might have their own policies on data.

Having said all that, it is a nice way to play around with various models.

Thank you, that pricing table is very very hidden (I even visited that section and missed it). It looks like they don't charge anything extra if you use GPT-4, for example, which is great, as I can point my code to them to use GPT-4 and then just switch models out as necessary.

> It looks like they don't charge anything extra if you use GPT-4

Yup, which makes it great to experiment without extra cost. It also should give you pause before you implement this in production, as it very much raises the question of what their income model is. Generally speaking, there are a few indicators that make me a bit hesitant to see this as anything more than a way to do some experimenting. Then again, a lot of that is true for making anything long term with the current AI offering.

I’ve been using it in production and it works great. Makes a world of open source models just as easy to use as OpenAI.

Here’s my Ruby gem for it. https://github.com/OlympiaAI/open_router