[KoolKat23]

Subscription costs per person going from $20 to $100, may be expensive and potentially unbudgeted expenditure, but it's hardly ruinous.

Many big organisations are already on per token pricing purely due to privacy concerns.

Once workflows are in place it becomes easier to see what level of "intelligence" is adequate. It no longer becomes important to necessarily keep that workflow on the next state of the art model if the new cheaper models does a good enough job given ongoing improvements. Gemma 4 I'd say is equivalent to 2.5 flash in my opinion/workflows.