Context
What happened: Banco Santander said on Oct. 19 that using smaller models for banking tasks cut AI serving costs in selected tests, and set a three-year framework for further work with NVIDIA.
Key numbers: About €0.03 per interaction, down from roughly €5. In an experimental customer service use case, costs fell more than 50%.
Why it matters: The bank linked lower serving costs to smaller task-specific models and NVIDIA Dynamo-Triton, and plans work on foundation models, AI agents, fraud and anti-money laundering capabilities under the framework.
Why it matters
The bank linked lower serving costs to smaller task-specific models and NVIDIA Dynamo-Triton, and plans work on foundation models, AI agents, fraud and anti-money laundering capabilities under the framework.