A number is worth as much as what stands behind it. Every figure we put in front of you traces to one of two things: published provider pricing, or your own traffic, measured. Nothing rests on a slide.

Grounded in published pricing

The cache economics are public. A cache read on frontier providers bills at about a tenth of a normal input token, and the write premium is a known multiple of the input price. From those rates and the shape of a multi-turn session, you can work out the saving on the repeated context before a single request is routed. That arithmetic is where the illustrative figures on this site come from, and we say so wherever one appears.

Measured on your own traffic

An estimate from published pricing is a starting point. The figure that matters is the one your own sessions produce. So before anything is agreed, we run a sample of your traffic through the same routing and caching we would use in production, set the result against what you pay today, and show you the difference on workloads you recognise.

You see that number before you commit. We do not quote a market size or a customer count, because neither tells you what you would save. The thing that does is your bill, before and after, on your own work.