← Benchmarks·RETENTION·NO PUBLIC BENCHMARK

Tokens (or AI calls) per active user

Usage intensity that drives both cost and value of AI features.

Formula
Total tokens (or model calls) / active users per period
Unit
count
Models
SaaS, Usage-based
Benchmark
Directional
ALLCapacity-planning and margin driver; the consumption variable behind inference COGS. Heavy-tailed (a few power users dominate cost) - track the DISTRIBUTION, not just the mean. Glean cited ~20T tokens/yr as scale color.
Sourcing: None (scale).
Omega Point BenchmarksRetention