Korva is a scroll of the internet’s best ideas.
The second cut came in coordination with OpenRouter, and together the two cuts took Luna's price down 10x in two weeks. Usage jumped 13x, flattened out at that level, and then went back to growing at the same rate it had before the jump.
The cuts landed while DeepSeek launched with a very good price and GLM also had a very good price. Luna is now used more than GLM on OpenRouter, and GLM had been one of the top three or four models by token volume.
It is the first time in an extremely long while that an OpenAI model has ranked among the top three to five models on the platform by token volume. Usage growing by more than the price fell is the pattern known as the Jevons paradox.
Why does cheaper often mean more total use?
The pattern is named for the economist William Stanley Jevons, who described it in his 1865 book The Coal Question. He noticed that more efficient steam engines made each unit of work cheaper, yet Britain's total coal consumption rose instead of falling. Cheaper power made factories, railways, and uses viable that had never been worth the cost before. Economists now call the broader effect a rebound.
How does OpenRouter pick which provider serves a request?
OpenRouter offers one API for many models and forwards each request to one of the providers hosting that model. Routing can weigh price, speed, and recent uptime, and it can fall back to another provider when one fails. Several companies can host the same open weight model, which is what lets traffic move between them.
Why do model prices keep falling so fast?
Epoch AI analyzed model pricing in 2025 and found that the price of reaching a fixed level of capability fell somewhere between 9x and 900x per year, depending on the capability threshold. Part of the drop comes from more efficient chips and better serving software. Another part comes from smaller models that are distilled from larger ones and match them on specific tasks.
More from 20VC with Harry Stebbings
