CoinWorld News, a SemiAnalysis report shows that internal large model token spending has accounted for 30% of total employee salaries, with per capita monthly consumption of nearly 5 billion tokens, and core contributors consuming over 100 billion tokens per month.


The sharp drop in actual usage costs is the key to reshaping the unit economics of the professional services industry.
Although the official price of Opus 4.7 is as high as $5 per million tokens for input and $25 per million tokens for output, due to the agent tasks having a 300:1 input-to-output ratio and over 90% prompt cache hit rate, the actual blended token cost is only $0.99 per million.
The combined acceleration of software and hardware is further compressing generation costs.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • 3
  • 1
  • Share
Comment
Add a comment
Add a comment
OvernightPositionPhobia
· 4h ago
90% cache hit rate + 300:1 input-output ratio, actual cost cut to $0.99, this is the real cost reduction and efficiency improvement.
View OriginalReply0
BlueGlassJelly
· 4h ago
Core contributors spend 100 billion every month, and it feels like in the future resumes will need to list token processing volume.
View OriginalReply0
YieldYuki
· 5h ago
Per capita 5B tokens, that number makes my wallet tighten.
View OriginalReply0