Live from production/Jun 29 to Aug 31/40 models/36 providers
The models teams run in production, and what they pay for them.
18.1%
of tokens on deepseek-v4-flash-0731
▲ 6.8pp
glm-5.3-flash, week over week
$0.76
blended cost per million tokens
67%
cut from the bill by caching