What type of study is this?

This is a Quantitative Study study.

October 1, 2025Open Access

Sustainability via LLM Right-sizing

View Full Paper

Ask AI

Bookmark

Authors

Empirical analysis assesses energy consumption and output quality across various tasks, suggesting sustainable model deployment.

GPT-4o shows superior performance but has a higher cost and environmental footprint, suggesting trade-offs in model selection.
Smaller models like Gemma-3 and Phi-4 performed reliably well, indicating they are viable for cost-efficient and locally deployable solutions.
The analysis identified three distinct model groups, revealing trade-offs between quality, control, and sustainability in model performance.
Task type significantly influenced model effectiveness, with conceptual tasks challenging most models while aggregation tasks fared better.

Haase et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark