AI-Powered, But Power-Hungry? Energy Efficiency of LLM-Generated Code

Lola Solovyeva, Sophie Weidmann, Fernando Castor · 2025

Read AI-Powered, But Power-Hungry? Energy Efficiency of LLM-Generated Code by Lola Solovyeva, Sophie Weidmann, Fernando Castor (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-02412.txt.

Description

Lola Solovyeva, Sophie Weidmann, and Fernando Castor measured the energy cost of code written by language models in a 2025 study. AI assistants now draft and complete code across the software industry. Reviews of their output check whether the code is right, but rarely whether it runs lean or wastes power.

The authors benchmark model-written programs for speed and energy use against human-written versions. Slow code is not a trivial flaw in systems that run at huge scale, where small waste multiplies into real emissions and bills. Their results show where models produce needlessly heavy code. The study pushes model evaluation beyond correctness, toward the full cost of running what the model writes.

Catalog record

CREATORLola Solovyeva, Sophie Weidmann, Fernando Castor
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE53 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-02412

Browse the full library · About the Tower of Babel