Exowatt beats frontier models on accuracy with SageMaker serverless model customization
Exowatt's fine-tuned edge models beat frontier LLMs at 49x lower cost using SageMaker
“achieving higher accuracy than frontier models at up to 49× lower cost”
Exowatt, a modular solar energy company, used Amazon SageMaker serverless model customization to build task-specific fine-tuned models for alarm triage on edge hardware with no GPU and limited connectivity. The resulting models outperform frontier models on the narrow task at up to 49x lower cost, with the pipeline able to produce certified, deployable models in days. This is a concrete data point in the growing case for domain-specific fine-tuned small models over frontier models for high-frequency, narrow operational tasks.