Aiven builds LiteLLM-based gateway, cuts AI token costs by 80%

Finland-based open source database provider Aiven says an internal LLM gateway built on LiteLLM has cut its AI token spend by roughly 80%. The gateway gives developers access to open-weight models with budgeting and access controls, while frontier models are reserved for planning work. Aiven plans to offer the gateway as a service to customers.

• Internal LiteLLM gateway cut token spend by about 80%
• Open-weight models do most work; frontier models handle planning
• Models run on rented GPUs and are self-hosted
• Service will target customers including Sophos, Avaya and Dojo