We kept noticing the same thing in the community: developers love DeepSeek V4 Pro. It's everywhere — in side projects, in production, in the tools people reach for first.

And looking at it, the relationship between its price and its usage seemed interesting— not the cheapest, not the priciest, yet consumed at a volume that didn't line up with cost the way you'd naively expect.

That made us wonder: what if we tallied every model ZenMux serves the same way — its standardized basket price against the real tokens it serves per day — and looked at the whole field at once? Where does the compute, and the money, actually flow?

So we did. This study is that tally: every text model on the platform, scored on two axes — what it costs and how much it's truly used — so the value frontier becomes something you can see rather than guess.

“Price is what you pay. Usage is what the market actually believes.”