Very fast drain of usage compared to last weeks
retained text
My usage drain compared to last week is insanely fast while at the same time I improved my process so that I use opus on medium instead on xhigh and as much as it can handoff onto agents it does, 80% of the time it uses haiku agents, other times it uses sonnet and a very low amount of times it uses opus on high. Now the kicker. Currently I'm at 33% usage while when I compare the tokens used, I should be around 14% (not accounting for the lower models being used). Cache hits and cache writes still have the same ratio as last week and my cache hits are still at 98%. I wonder if anyone else is seeing the same: usage limits drain faster then last week? Edit, I was asked for the numbers: **Last week (Oct 3 – Oct 10, 7 days), this was 100% usage on max 20:** - Output tokens: 50.8M - Cache writes: 150.4M - Cache reads: 12.2B - Fresh (uncached) input: 113k - API calls: 56.4k - Sessions: 667 - Cache hit rate: 98.8% **This week so far (first 24 hours) this is 40% usage on max 20** - Output tokens: 10.2M - Cache writes: 32.6M - Cache reads: 2.1B - Fresh input: 26k - API calls: 12.9k - Sessions: 140 - Cache hit rate: 98.5% According to the numbers my usage should be around 20% 90% of tokens are opus 5.5 medium, 2% opus 5.5 high or xhigh and the rest is sonnet/haiku.
document fetched 2026-10-12 00:36 · canonical https://reddit.com/r/ClaudeCode/comments/1x2xak4/very_fast_drain_of_usage_compared_to_last_weeks