A per-token model roughly aligns with the providers' costs, and it is an objective measure, so it seems a reasonable way to charge. I see posts all the time on

A per-token model roughly aligns with the providers' costs, and it is an objective measure, so it seems a reasonable way to charge. I see posts all the time on HN about which models from which providers offer the most bang-for-the-buck, and how to minimize token usage and still get optimal results, so it appears that competition is working.