Improving the Energy-Efficiency of the Code Generated by LLMs through Effective Prompting

Ritika Rekhi; Bing Zhang; Md Arman Islam; Jaya Krishna Pasham; Yeswanth Chitturi; Akshay Paramesha; Isha Valiveti; Asif Imran

cs.SEOct 1, 2026

Improving the Energy-Efficiency of the Code Generated by LLMs through Effective Prompting

Authors: Ritika Rekhi, Bing Zhang, Md Arman Islam, Jaya Krishna Pasham, Yeswanth Chitturi, Akshay Paramesha, Isha Valiveti, Asif Imran,

Organizations: University at Buffalo, Buffalo, NY, USA · National Center for Supercomputing Applications, Urbana, IL, USA · IBM Research, Yorktown Heights, NY, USA

Abstract

As AI-assisted programming becomes increasingly mainstream, the environmental impact of AI-generated software has emerged as an important consideration. This motivates evaluating LLM-generated code beyond functional correctness by considering execution efficiency and energy consumption. However, despite substantial advances in code generation, frontier LLMs are rarely evaluated based on the energy efficiency of the code they produce. In this work, we conduct a comprehensive evaluation of 21 prompting strategies for energy-efficient code generation and identify 8 strategies for evaluation across 10 widely used open-weight and proprietary LLMs. We evaluate their effectiveness for both Python and C++ code generation relative to a baseline prompt. Across the evaluated models, the selected prompting strategies achieved energy reductions of up to 25% for Python and 17% for C++ code generation. At the model level, Python energy reductions reached up to 50% for Granite-4.0-H-Small, 39% for Claude 4.5 Haiku, and 28% for MiniMax M3, while C++ reductions reached up to 57% for Granite-4.0-H-Small and 6.4% for Qwen3-Coder-480B-A35B-Instruct. These results demonstrate that prompting strategies can substantially influence the energy consumption of LLM-generated code, although their effectiveness varies across models and programming languages. Our findings highlight the importance of incorporating energy efficiency into the evaluation and optimization of LLM-based code generation and provide practical insights into designing prompts for more sustainable AI-assisted programming.

Figures & tables

Model Details			Model Configuration
Model Name	Model Weights	Knowledge Cut-off	API Provider	API Type Used	Temp.	Top_p	Top_k	Reasoning / Thinking Configuration
Claude 4.5 Haiku	Closed	Feb. 2025	Anthropic	Batch	∗ 1	∗ 1	∗ 0	∗ Thinking: Disabled
Gemini 3.8 Flash	Closed	Mar. 2026	Google Vertex AI	Batch	N/A	N/A	N/A	∗ Thinking Level: Medium
GPT-5.6 Terra	Closed	Feb. 2026	OpenAI	Batch	N/A	N/A	N/A	∗ Reasoning Effort: Medium
GPT-OSS-120B (116.8B total / 5.1B active)	Open	Jun. 2024	Fireworks AI	Batch	1	1	∗ 0	Reasoning Effort: High
Gemma 4 26B A4B IT (25.2B total / 3.8B active)	Open	Jan. 2026	Fireworks AI	Batch	1	0.95	64	Thinking: Enabled

TABLE I : Overview of the LLMs under study along with their respective configurations used for the study. An asterisk (

*

) indicates the default value used by the model provider, while N/A denotes a configuration that is not applicable to the model.

Fig. 1 : Code generation workflow applied to each prompting strategy across both evaluation stages.

Fig. 2 : Code evaluation workflow for the generated Python and C++ programs.

Fig. 3 : Fine-grained energy measurement process for Python and C++ programs.

Fig. 4 : EffiBench prompt structure used as the baseline.

Prompt Number	Conversation Type	Source	Prompt Type	Change Scope	Additions Made to EffiBench Prompt
$\star$ Prompt 0 (EffiBench Prompt)	SingleTurn	[ 14 ]	Zero-Shot	-	-
Prompt 1	SingleTurn	[ 11 ] [ 13 ] [ 27 ] [ 28 ]	Zero-Shot	User Prompt	Energy-Efficiency Keyword
Prompt 2	Single-Turn	-	Zero-Shot	User Prompt	Energy-Efficiency Keyword and Definition
Prompt 3	Single-Turn	-	Few-Shot	User Prompt	Energy-Efficiency Keyword, General Efficiency Guidelines
$\star$ Prompt 4	Single-Turn	-	Persona-Based + Zero-Shot	User Prompt	General Efficiency Guidelines, Language-Specific Efficient Practices
Prompt 5	Single-Turn	-	Persona-Based + Zero-Shot	System Prompt	Expert Competitive Programmer Persona, Energy-Efficiency Definition

TABLE II : Summary of 21 prompting strategies and the EffiBench prompt evaluated in the initial analysis. Strategies marked with

\star

and highlighted in green were selected for the final evaluation for Python and C++ code generation across our extended EffiBench dataset and all LLMs under evaluation.

Panel of Figure 7 — Fig. 5 : Finalized prompting strategies for Python and C++ code generation. The top panel presents the five single-turn prompting strategies, while the bottom panel presents the three multi-turn prompting strategies. Both are applied across the extended EffiBench dataset and evaluated across all LLMs under study.

Fig. 6 : Average change in energy consumption relative to the baseline prompt across 110 python problems and 21 prompting strategies, aggregated across the 3 LLMs evaluated in the initial analysis.

	Algorithms
Difficulty	Backtracking	BFS	Binary Search	Bit Manip.	DFS	Divide & Conquer	DP	Greedy	Sliding Window	Sorting	Two Pointers	Total
Easy	1	0	13	21	0	2	7	25	7	47	24	110
Medium	17	21	48	33	24	3	96	98	31	83	33	333
Hard	3	6	24	7	2	4	45	18	8	17	0	94
Total	21	27	85	61	26	9	148	141	46	147	57	537

TABLE III : Distribution of benchmark problems solved by all 10 evaluated LLMs across difficulty levels and algorithm categories for Python code generation.

Fig. 7 : Average percentage energy reduction achieved by the 8 prompting strategies relative to the baseline prompt for Python code generation. Negative values indicate lower energy consumption than the baseline prompt.

Panel of Figure 11 — Fig. 8 : Impact of prompting strategies on the energy consumption of Python code generation relative to the baseline prompt across problem difficulty levels and algorithmic categories, aggregated across all evaluated models.

	Algorithms
Difficulty	Backtracking	BFS	Binary Search	Bit Manip.	DFS	Divide & Conquer	DP	Greedy	Sliding Window	Sorting	Two Pointers	Total
Easy	1	0	20	19	0	1	5	31	4	56	26	121
Medium	9	9	55	40	13	3	105	132	37	97	43	373
Hard	0	7	22	11	2	1	66	15	8	21	1	113
Total	10	16	97	70	15	5	176	178	49	174	70	607

TABLE IV : Distribution of benchmark problems solved by all 10 evaluated LLMs across difficulty levels and algorithm categories for C++ code generation.

Fig. 9 : Average percentage energy reduction achieved by the 8 prompting strategies relative to the baseline prompt for C++ code generation. Negative values indicate lower energy consumption than the baseline.

Panel of Figure 14 — Fig. 10 : Impact of prompting strategies on the energy consumption of C++ code generation relative to the baseline prompt across problem difficulty levels and algorithmic categories, aggregated across all evaluated models.

Panel of Figure 15 — Fig. 11 : Overall impact of 8 prompting strategies Python and C++ code generation for their respective evaluation datasets aggregated across all models.

LLM Code Generation

Large language model code generation studies how models translate instructions and context into executable programs and how to make them reliable.

See all 308 papers

An Initial Exploration of Contrastive Prompt Tuning to Generate Energy-Efficient Code
Mar 3, 2026Sophie Weidmann, Fernando CastorCode Generation LLM Code Generation
University of Twente, Enschede, The Netherlands
Green Code or Costly Code? Evaluating the Energy Costs of Iterative LLM-Based Code Optimizations
Oct 23, 2026Sadra Emami, Benjamin Weigell, Mohammad Reza Mousavi +3Large Language Model-Guided Optimization Code Generation
University of Augsburg, Augsburg, Germany · King’s College London, London, United Kingdom
Beyond the Need for Speed: Energy-Aware Code Generation via Simulation-Guided Reinforcement Learning
Jul 6, 2026Saurabhsingh Rajput, Tushar SharmaRL for Code Generation Reinforcement Learning
Dalhousie University, Halifax, Canada

Mar 3, 2026J/K move · Enter open · S save

An Initial Exploration of Contrastive Prompt Tuning to Generate Energy-Efficient Code

Sophie Weidmann, Fernando Castor

University of Twente, Enschede, The Netherlands

Code Generation LLM Code Generation Prompt Tuning

Open paper AskPDF arXiv

Although LLMs are capable of generating functionally correct code, they also tend to produce less energy-efficient code in comparison to human-written solutions. As these inefficiencies lead to higher computational overhead, they are in direct conflict with Green Software Development (GSD) efforts, which aim to reduce the energy consumption of code. To support these efforts, this study aims to investigate whether and how LLMs can be optimized to promote the generation of energy-efficient code. To this end, we employ Contrastive Prompt Tuning (CPT). CPT combines Contrastive Learning techniques, which help the model to distinguish between efficient and inefficient code, and Prompt Tuning, a Parameter-Efficient Fine Tuning (PEFT) approach that requires only a fraction of the cost of traditional fine tuning. This study evaluates CPT on Python, Java and C++ coding problems across three different models to provide a comprehensive evaluation. The method achieves consistent improvements in code accuracy for two models but efficiency gains vary by model, language and task complexity, indicating that improvements are not uniformly reliable.

Abstract

LLM Code Generation

Explore similar work

An Initial Exploration of Contrastive Prompt Tuning to Generate Energy-Efficient Code

Green Code or Costly Code? Evaluating the Energy Costs of Iterative LLM-Based Code Optimizations

Beyond the Need for Speed: Energy-Aware Code Generation via Simulation-Guided Reinforcement Learning