Back to AI News
arxiv

Toward Sustainable Agentic AI: Carbon Tradeoffs Across Iterative Self-Refinement and Tool-Augmented Coding

A new arXiv paper presents two empirical studies estimating the energy and carbon footprint of agentic language-model systems. In iterative self-refinement on MATH, three refinement steps closed a 16-percentage-point accuracy gap to within one point but nearly tripled carbon versus a cheap single-pass baseline; tool-augmented coding on SWE-bench Lite showed GPT-5.3-codex at 81.2% resolution, ahead of Claude Opus 4.6 at 68.9% and GPT-4.1-mini at 25.0%, while GPT-4.1-mini had the lowest estimated carbon and cost per resolved task. The authors note that carbon estimates cover output-token inference only and exclude prompt tokens and tool execution, making them lower bounds on the full agentic footprint.

AI Brief

It quantifies a concrete tradeoff between agentic accuracy gains and inference carbon, helping teams evaluate agent configurations beyond performance alone.

arXiv:2611.12128v1 Announce Type: new Abstract: Agentic language model systems can improve performance either by iterating on their own outputs or by using external tools in grounded environments. While these strategies can reduce error rates and substitute for larger models in some settings, their environmental cost remains relatively underexplored. We present two complementary empirical studies that estimate the energy and carbon footprint of agentic AI across settings of increasing complexity. Study~1 evaluates iterative verbal self-refinement: GPT-4.1-mini at $k = 1$--$5$ steps versus a GPT-4.1 single-pass baseline on two MATH benchmark splits. Study~2 evaluates tool-augmented agentic coding: GPT-4.1-mini, Claude Opus~4.6, and GPT-5.3-codex on SWE-bench Lite, where agents operate inside Docker-isolated environments with access to code execution and file editing tools. In Study~1, $k=3$ refinements close a 16-percentage-point accuracy gap to within 1~point, but nearly triple the carbon footprint relative to the cheap single-pass baseline while remaining below the GPT-4.1 single-pass baseline under the current token-based carbon scenario. In Study~2, GPT-5.3-codex achieves the highest resolution rate in this sample (81.2\%), ahead of Claude Opus~4.6 (68.9\%) and GPT-4.1-mini (25.0\%), but resolved-task-normalized efficiency remains mixed: GPT-4.1-mini is lowest on estimated carbon and cost per resolved task, while GPT-5.3-codex delivers substantially higher task success. All carbon estimates cover output-token inference only; prompt tokens, which exceed 99\% of tokens in Study~2, and tool execution are not charged, so the figures are lower bounds on the full agentic footprint.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!