Now that's interesting: if I use our librechat instance, with pay-as-you-go API access to OpenAI, and I send a request to gpt-5.1, I get the following answer:

> I’m based on the GPT‑4.1 family of OpenAI models

But if I look at the dashboard of OpenAI, it counts this as a request to the 5.1 model.

So I pay for a 5.1 model, but get routed to a 4.1? Anybody any idea what is going on here?