Researchers have uncovered a significant environmental cost associated with our increasing reliance on sophisticated AI systems. A new study published on June 19, 2025, in Frontiers in Communication reveals that reasoning-enabled AI models can emit up to 50 times more carbon dioxide than their simpler counterparts when answering identical questions.
The research team, led by Maximilian Dauner from Hochschule München University of Applied Sciences, evaluated 14 different large language models (LLMs) ranging from 7 to 72 billion parameters. They tested these models on 1,000 benchmark questions across diverse subjects including mathematics, history, philosophy, and abstract algebra.
The study found that reasoning models generated an average of 543.5 'thinking tokens' per question, compared to just 37.7 tokens for concise models. These additional computational steps directly translate to higher energy consumption and carbon emissions. The most accurate model tested was the reasoning-enabled Cogito model with 70 billion parameters, which achieved 84.9% accuracy but produced three times more CO2 than similarly-sized models generating more concise answers.
"Currently, we see a clear accuracy-sustainability trade-off inherent in LLM technologies," explained Dauner. "None of the models that kept emissions below 500 grams of CO2 equivalent achieved higher than 80% accuracy."
The subject matter of questions also significantly impacted emissions. Questions requiring complex reasoning, such as abstract algebra or philosophy, led to up to six times higher emissions than straightforward topics like high school history.
The researchers highlighted that users can control their AI carbon footprint through thoughtful choices. For example, DeepSeek's R1 model (70 billion parameters) answering 600,000 questions would create CO2 emissions equal to a round-trip flight from London to New York. Meanwhile, Alibaba's Qwen 2.5 model (72 billion parameters) could answer about 1.9 million questions with similar accuracy while generating the same emissions.
"If users know the exact CO2 cost of their AI-generated outputs, they might be more selective about when and how they use these technologies," Dauner concluded. The researchers hope their work will encourage more informed and environmentally conscious AI usage as these technologies become increasingly embedded in our daily lives.