Exowatt fine-tunes Qwen3 models for edge and cloud tasks at lower cost
Exowatt, a solar energy provider for AI data centers, built a reusable AI pipeline using Amazon SageMaker’s serverless model customization. The company fine-tuned Qwen3-1.7B for edge alarm triage and Qwen3-14B for cloud energy-market analysis, achieving higher accuracy than frontier models at up to 49× lower cost per decision. Both models run autonomously, citing specific data rows to justify…
Key points
- Fine-tuned Qwen3-1.7B model outperformed frontier models on edge alarm triage with 0.911 macro-F1 at $0 cost per 1,000 decisions
- Qwen3-14B model achieved 0.975 macro-F1 on cloud energy-market analysis, costing $0.60 per 1,000 answers vs. $29.26 for Claude Opus 4.7
- Exowatt reduced model development time from months to days by using SageMaker’s serverless fine-tuning and avoiding GPU provisioning
Exowatt beats frontier models on accuracy with SageMaker serverless model customization
AWS Machine Learning Blog · 27 October 2026
Loading the full article…
This text was published by AWS Machine Learning Blog and written by Purna Sanyal. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Hardware & Compute
All →- Oracle Cloud to add Google TPUs for AI training and inference in 2027 · 2 src
- Oracle adds RDMA and multiplanar failover to Acceleron for AI workloads · 1 src
- NVIDIA and Oracle expand OCI Superclusters to multi-gigawatt Vera Rubin scale · 3 src
- Nvidia and TSMC both hit records but stocks fall amid supply risks · 5 src
- Modders bypass Nvidia’s AI performance limits on RTX 20, 30, 40-series GPUs · 2 src
Comments
via GitHub Discussions