How exactly does OpenRouter handles rate/tokens/requests limits?

[jade cobalt]

I’ve got a question about OpenRouter limits

If I make a request directly to OpenAI's gpt-4o, because I'm in tier 2 I currently get max 5000 requests per minute and 450k tokens per minute and unlimited request per day. These limits apply at the organization level, so using multiple API keys doesn’t make a difference—it’s all counted together.

Now, for OpenRouter (if I use openai/gpt-4o-2024-11-20), I saw in the docs that limits are tied to credits, with something like 1 request per credit per second. So, if I have $5 in credits, that would mean 5 requests per second, right?

But what about token limits? Are there any, or is it unlimited as long as you have enough credits?

[rose garnet]

OpenRouter only has requests per second limits for paid models, based on credits as you described, no token limits.

[jade cobalt]

Great, thank you

[rose garnet]

If the actual provider enforces different rate limits, e.g. Google for their experimental models, OpenRouter will forward those rate limit errors and there is nothing OpenRouter can do about this.
See e.g. here -> #1314643693009113100 message

[jade cobalt]

Ok, but for OpenAI models, at least for now, there's no token limits in OpenRouter , right?

[rose garnet]

But this should be the exception, not the rule.

[rose garnet]

Correct.