Plain inference is profitable actually, that’s why there are a hundred inference providers on OpenRouter who compete by undercutting each other. The labs however aren’t profitable because training the models is a huge drain.
As others have pointed out, the compute cost of inference is only one small part of the puzzle. All the frontier model providers - which OpenRouter gives you access to - are massively raising their prices in desperate bids to recoup the cost of model generation in the first place.
There’s really no hiding from the token apocalypse unless you’re running a model on your own hardware.
Plain inference is profitable actually, that’s why there are a hundred inference providers on OpenRouter who compete by undercutting each other. The labs however aren’t profitable because training the models is a huge drain.
As others have pointed out, the compute cost of inference is only one small part of the puzzle. All the frontier model providers - which OpenRouter gives you access to - are massively raising their prices in desperate bids to recoup the cost of model generation in the first place.
There’s really no hiding from the token apocalypse unless you’re running a model on your own hardware.