The GPU-Hour Index tracks the median $/GPU-hr across four standardized SKUs. The Token Index tracks median output $/Mtoken across a curated set of frontier models, both open and closed weights. Full methodology and constituents below.
For each index we compute annualized drift and volatility from hourly log returns over the last 30 days. The forward point at horizon T is spot · exp(drift · T / 365); the band is ±1.96σ·√(T/365). Indicative of where the market has been heading — not a price you can transact at.
On-demand $/GPU-hr across four standardized GPU configurations, taken cross-provider and rolled up into a single number.
Output token price across five open-weights frontier models, averaged into a single $/Mtoken figure.
Output token price across three closed frontier models, averaged into a single $/Mtoken figure.
H100 SXM5 80GB with NVLink interconnect, 8-GPU node
1 million output tokens at the published on-demand rate per constituent model
Current values and methodology are available at /api/v1/tci, cached publicly for five minutes.