That said, I don't understand the people who use something a full agentic backbone with expensive models like Claude Opus with OpenRouter because that 5% surcharge is meaningful at that level of cost instead of going with the source API providers. But people are clearly doing it, and it's pure revenue.

1. By far the lowest friction way to support and try out all the models.

2. They offer billing caps! Most model providers still don't do this [EDIT: maybe they do, see reply comment], but if you're going to run anything in public it's very useful to have hard limits so it doesn't cost you $1m overnight because someone started abusing it.

3. Their rankings are one of the more interesting signals for which models are popular, despite their flaws (most OpenAI and Anthropic users don't go via OpenRouter, it's currently not possible to tell the difference between many users switching v.s. one "whale" changing their preferred model)

Given how API costs are becoming meaningful for a lot of companies now, having a provider like OpenRouter to help measure your spend and easily experiment with and switch providers feels like a valuable service.

Are tech companies FOMOing so hard that they're now all running AI venture arms themselves instead of you know, developing their own products? Except for NVIDIA who needs to keep pumping the bubble I didn't expect the others.

After things begin to settle down, we'll probably see a consolidation of both frontier and open-source models - and then OpenRouter will become less useful, because that 5% overhead is well worth it when you want to try 20 models from 10 labs, but harder to stomach when you only need 5 models from 2 providers, and each of those providers has its own API knobs that you can tune to make things even cheaper.