The benchmark for the cost of AI capability. One fixed methodology, tracked over time, measuring what a unit of standardized intelligence costs across open and closed models.
Each index is set to 100 at launch and moves as prices change: a reading above 100 means AI capability is cheaper than at launch (more intelligence per dollar); below 100 means dearer. The three baskets let a buyer see the whole market, the open-weight market, and the frontier market separately, and the premium between them.
A tool built on the same data: for a given job, the best-value model on a capability-adjusted basis. Pick the job; the ranking reweights quality against cost.
| # | Model | Quality | $/M | Source | Job score |
|---|
Methodology: quality-adjusted cost = blended price (1:3 input:output) ÷ capability score, equal-weighted across a fixed basket, indexed to 100 at launch. Baskets reviewed periodically. The index launches equal-weighted for transparency and transitions to consumption-weighting as Cambent's routing volume yields representative usage data.
Data: 4 live prices with per-model provenance, 10 manual and clearly labeled until each source is verified. Quality scores are illustrative composites pending the live benchmark layer.