Hermes research work quickly used up z.ai’s weekly allowance
Hermes worked well when used lightly through a z.. After the workload expanded to and other tasks, the entire weekly allowance ran out much sooner than expected. The z.ai account was used only for Hermes, so another service did not consume the allowance.
Hermes appeared to use many tokens, but the setting or usage pattern responsible was not identified. A practical way to reduce token use and manage the is still needed.
Key points
- Light Hermes use did not cause a problem with the limit.
- and other tasks quickly consumed the full weekly z.ai allowance.
- The z. was used only for Hermes.
- No proven method or concrete best practice was provided.