us. So I think what's happening is when you look at the frontier models, they have large data centers to cater for the average user. 

Businesses are paying the tax for that. They're paying the tax for the agents to know every piece of data known to the consumer. 

humankind and then also to service people that are, you know, maybe using it for college, maybe using it for for home life. 

So your companies now can't sustain that cost. And so what they're doing is they're bringing it internal and what a small language model will do is it'll fit on two GPUs, two pro 6000 GPUs. 

So if you think about that and you think about the cost per token, if you were running those those gpus and you were to divide that token cost into those, It comes down to under 35 cents per million tokens. 

So that's something that's bearable for a company. You