The Environmental Cost of AI's Reasoning Power

A recent study highlights the environmental challenge posed by chat-based generative AI, revealing that these systems can emit significantly more carbon when processing complex queries compared to simpler ones. The study focused on 14 large-language models, such as DeepSeek and Cogito, examining their response to 1,000 benchmark questions.

Researcher Maximilian Dauner from Hochschule München University of Applied Sciences pointed out that models capable of reasoning produce notably higher emissions. Specifically, reasoning models emit up to 50 times more CO₂ than models that produce concise responses. The substantial energy consumption stems from the intricate reasoning processes these models undertake.

Findings indicate a trade-off between accuracy and environmental impact. Although reasoning models like Cogito achieved nearly 85 percent accuracy, their emissions were markedly higher than concise models. The need for balancing accuracy with sustainability is crucial for advancing environmentally conscious AI technologies.

(With inputs from agencies.)