# Exowatt deploys specialized models for offline alarms and energy analysis

_Published Tuesday, October 27, 2026 at 1:09 PM EDT · AI, Infrastructure · Latest · Tier 2 — Notable_

![Exowatt deploys specialized models for offline alarms and energy analysis — Primary](https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/10/19/ML-22072-featured-image.png)

Exowatt has deployed fine-tuned Qwen models for solar-system alarm triage and energy-market analysis, AWS says. Its Qwen3-1.7B alarm model runs on existing site CPUs without a GPU or internet connection; its Qwen3-14B market model serves through Amazon Bedrock.

AWS reports macro-F1 classification scores of 0.911 across 320 alarm tests and 0.975 across 2,000 market tests. Both models cite input data, and unsupported claims are removed before reaching operators. The results concern narrow operational tasks, with alarm training examples drawn from generated site scenarios and certified paraphrases.

## Sources

- [Artificial Intelligence](https://aws.amazon.com/blogs/machine-learning/exowatt-beats-frontier-models-on-accuracy-with-sagemaker-serverless-model-customization/)

---
Canonical: https://techandbusiness.org/newswire/NM-QosWThgJgkxF5-q3XO_
Published: 2026-10-27T17:09:16.727Z
Story chronology: 2026-10-27T16:50:55.000Z
Retrieved: 2026-10-27T19:38:10.873Z
Publisher: Tech & Business (techandbusiness.org)