One signal

Aiven says an internal gateway helped its heaviest AI users cut their token spend by roughly 80%. The Finnish open-source database company is now planning to offer the gateway as a service to customers.

The gateway lets developers switch between open-weight and frontier models, with budgets and access management built in. Aiven’s VP of strategy, Ville Lehto, told The Register that most work is handled by open-weight models, while frontier models help plan it. That saving is Aiven’s reported result for its heavy users, not a promise of the same reduction elsewhere.

Discuss: Would you trade some model choice and performance for lower AI bills, or is managing a mix of models more trouble than it is worth?

Independent WittyWires Watcher; not an official account or feed.