Based on the latest communication about pricing and token consumption, my understanding was that using auto mode does not count towards your monthly credit allocation for accessing premium models. However, I’ve noticed that some of my auto/default usages are consuming tokens while others are not. I also observed that token consumption occurs specifically when auto mode uses the Sonnet 4 model (identifiable by the Sonnet-like responses and tool calls).
Questions:
Has anyone else noticed this behavior?
Isn’t auto mode supposed to be totally free, even when it automatically allocates Claude models?
Is there any way to disable the Sonnet model in auto mode to prevent token consumption?
Auto/default does not count towards usage limits (regardless of the model that it ends up routing to), but we still show the token counts for them (its actually looks buggy that any of them show up as 0); we’ll take a look to see what’s going on there.
But as you can see in my monthly usage summary, it is being counted towards my $20 included usage, and I’ve noticed that when usage entries have token consumption ≠ 0, the token consumption is being attributed to the default (auto) row.
We still show the default/auto tokens and api cost there to give an indication of the value gotten from it, but it doesn’t count towards the usage limit.
We’re pushing out an update soon to make the dashboard clearer there that the default model api price and tokens isn’t counted towards usage limits.
