Spend over time
By model, split by what you paid for
Efficiency
- Cache hit rate
- –
of context tokens served from cache
- Cost per prompt
- –
- Avg context per message
- –
input + cache tokens sent each turn
- Peak context
- –
- Cache writes at 1h rate
- –
1h writes cost 60% more than 5m
- Messages over 150k
- –
context this large costs the most per turn
Messages by context size
When you spend
local time
lessmore