One comparable unit
The TKX scores tokens across providers on quality, context, and cost — producing one comparable value and a fair exchange rate between models. About the TKX →
Cambent lets enterprises buy, sell, and convert AI capacity across labs — priced against one comparable standard, the capability-adjusted Cambent Token Index (TKX). Use the best model for every job. Recover what you don't use. Stay flexible as models evolve.
cambent — from cambiare, the old word for exchange.
The best model for reasoning isn't the best for code, and this quarter's leader may be next quarter's runner-up. Enterprises now run many models across many providers — and need to shift work and spend between them as capabilities evolve. But the plumbing is missing: no common unit to compare what a token from one provider is worth against another, no way to convert capacity between labs, and no market to recover what goes unused. That friction costs everyone — buyers and providers alike. Cambent exists to remove it.
The TKX scores tokens across providers on quality, context, and cost — producing one comparable value and a fair exchange rate between models. About the TKX →
Turn capacity you hold with one provider into usable capacity with another, cleared through the market at the index rate. How conversion works →
Post surplus, buy at or below index, recover stranded spend, and route every job to the best model for it. Explore the platform →
The industry's largest AI relationships are being repriced from compute-hours to tokens. The token is now the unit of account — before a standard exists.
Enterprises are burning annual AI budgets in a single quarter. Routing each job to the cheapest capable model is the new default — and routing demands a comparable unit.
The world's largest exchanges announced compute futures; major banks are exploring GPU trades. All of it is on hardware. The token layer is still open.