Hardware cost comparison
The table shows the lowest cost per million output tokens on each curve, at its highest measured throughput. Your cost increases when utilization falls. Estimated offers use the cheapest listed on-demand region for each cloud and GPU configuration.
No eligible alternative cloud prices are available for these measured configurations.
Estimates assume identical throughput for the same GPU configuration across clouds. Listed prices do not guarantee capacity. Measured pricing modes and regions are shown only where recorded by the source; estimates use on-demand pricing.