AI Reasoning Models Generate Up to 50 Times More CO₂ Emissions Than Concise Models, Study Finds

Certain AI prompts generate 50x more CO₂ than others

In recent years, researchers and climate advocates have been ringing the alarm about artificial intelligence's impact on the environment. Advanced and increasingly popular large language models (LLMs) -- such as those offered by OpenAI and Google -- reside in massive data centers that consume significant amounts of electricity and water to cool servers. Every time someone types a question or phrase into one of these platforms, the energy used to generate a response produces a measurable amount of potentially harmful CO₂. But, according to a new research published in Frontiers in Communication, not all of those prompts leave have the same environmental impact. Not even close. The study looked at 14 different LLMs, each varying in the size of their training data, and evaluated their performance using a standardized set of 500 questions across different subject areas. Each model generates a certain number of "thinking tokens" per query, and those tokens correlate with CO₂ emissions. When the researchers compared the responses, they found that more complex "reasoning models" -- which have larger training sets and take longer to process and respond -- produced significantly more CO₂ than smaller, more efficient "concise models." In some cases, reasoning models generated up to 50 times the emissions of their more concise counterparts. Aside from the models themselves, the amount of CO₂ generated by prompts also varied based on subject matter. More complex or open-ended questions, such as those involving advanced algebra or philosophy, tended to produce a larger carbon output than simpler prompts, like high school history questions. These findings shed further light on the often-overlooked ways AI models contribute to soaring energy consumption. Related: [AI will require even more energy than we thought] "The environmental impact of questioning trained LLMs is strongly determined by their reasoning approach, with explicit reasoning processes significantly driving up energy consumption and carbon emissions," Maximilian Dauner, PhD student at Hochschule München University of Applied Sciences and paper author, said in a statement. Reasoning models -- sometimes called "thinking models" -- refer to large LLMs optimized for solving more complex tasks that require logic, step-by-step breakdowns, or detailed instructions. These models often go by different names. At OpenAI, for example, GPT-4o and GPT-4o-mini are considered "generalized" models, while versions like o1 and o3-mini are classified as reasoning models. Reasoning models employ what some LLM researchers call "chain-of-thought" processing, allowing them to respond more deliberately than generalized models, which prioritize speed and clarity. The end goal is for reasoning models to generate more human-like responses. The most obvious by-product of that, for anyone who has used these, is that reasoning models take longer to generate answers. The researchers found that the reasoning models generated significantly more tokens, which correlate with CO₂ emissions, than the more concise models. (Tokens refer to words or parts of words that are converted into numerical representations the LLM can understand.) The testing occurred in two phases. In the first phase, researchers asked the same multiple-choice questions to models. The next, free response phrase, had the models provide written responses. On average, reasoning models generated 543.5 tokens per question, compared to just 37.7 tokens for concise models. The most accurate reasoning model they examined, called "Cogito," produced three times as much CO₂ as similarly sized models optimized for concise responses. "From an environmental perspective, reasoning models consistently exhibited higher emissions, driven primarily by their elevated token production," the researchers write in the paper. While the difference in emissions per individual prompt might seem marginal, it can make a real difference when scaled up. The researchers estimate that asking DeepSeek's R1 model to answer 600,000 questions would generate roughly the same amount of CO₂ as a round-trip flight from London to New York. By comparison, you could ask the non-reasoning Qwen 2.5 model three times as many questions before reaching the same level of emissions. Overall, the researchers say that their findings highlight a fundamental trade-off between LLM accuracy and environmental sustainability. "As model size increases, accuracy tends to improve," the researchers said. "However, this gain is also linked to substantial growth in both CO₂ emissions and the number of generated tokens." The findings come amid a fierce global race among tech companies to develop increasingly advanced AI models. Over the past year alone, Apple has announced plans to invest $500 billion in manufacturing and data centers over the next four years. Similarly, Project Stargate -- a joint initiative by OpenAI, SoftBank, and Oracle -- has also pledged to spend $500 billion on AI-focused data centers. Researchers warn that this surge in infrastructure could place additional strain on already overburdened energy grids. AI applications, in particular, play an outsized role in the energy consumption of newer data centers. A recent report in the MIT Technology Review notes that starting around 2017, data centers began incorporating more energy-intensive hardware specifically designed for complex AI computations. Energy use surged after that. The Electric Power Research Institute (EPRI) estimates that data centers supporting advanced AI models could account for up to 9.1 percent of the United States' total energy demand by the end of the decade -- up from approximately 4.4 percent today. Companies are scrambling to find new ways to meet this growing energy demand. Meta, Google, and Microsoft have all partnered with nuclear power plants to generate more electricity. Microsoft, one of OpenAI's primary partners, even signed a 20-year agreement to source energy from the Three Mile Island nuclear facility in Pennsylvania, a site once known for the worst reactor accident in U.S. history. Meta is also making major investments in geothermal technology as a less fossil fuel-intensive way to generate power. Others, like OpenAI CEO Sam Altman, who has said the coming age of AI will require an "energy breakthrough" are investing in experimental nuclear fusion. These investments may help companies make progress, but recent research indicates it's almost certain that more fossil fuels -- namely natural gas -- will be needed to fully meet AI's massive energy demand. Related: [The future of AI is even more fossil fuels] That may all sound daunting, but the researchers comparing different types of models say their findings could help empower everyday AI users to take steps to reduce their own carbon impact. If users understand how much more energy-intensive reasoning models are, they may choose to use them more sparingly and rely on concise models for general everyday tasks, such as web searches and answering basic questions. "If users know the exact CO₂ cost of their AI-generated outputs, such as casually turning themselves into an action figure, they might be more selective and thoughtful about when and how they use these technologies," Dauner said.

Thinking AI models emit 50x more CO2 -- and often for nothing

No matter which questions we ask an AI, the model will come up with an answer. To produce this information - regardless of whether than answer is correct or not - the model uses tokens. Tokens are words or parts of words that are converted into a string of numbers that can be processed by the LLM. This conversion, as well as other computing processes, produce CO emissions. Many users, however, are unaware of the substantial carbon footprint associated with these technologies. Now, researchers in Germany measured and compared CO emissions of different, already trained, LLMs using a set of standardized questions. "The environmental impact of questioning trained LLMs is strongly determined by their reasoning approach, with explicit reasoning processes significantly driving up energy consumption and carbon emissions," said first author Maximilian Dauner, a researcher at Hochschule München University of Applied Sciences and first author of the Frontiers in Communication study. "We found that reasoning-enabled models produced up to 50 times more CO2 emissions than concise response models." 'Thinking' AI causes most emissions The researchers evaluated 14 LLMs ranging from seven to 72 billion parameters on 1,000 benchmark questions across diverse subjects. Parameters determine how LLMs learn and process information. Reasoning models, on average, created 543.5 'thinking' tokens per questions, whereas concise models required just 37.7 tokens per question. Thinking tokens are additional tokens that reasoning LLMs generate before producing an answer. A higher token footprint always means higher CO2 emissions. It doesn't, however, necessarily mean the resulting answers are more correct, as elaborate detail that is not always essential for correctness. The most accurate model was the reasoning-enabled Cogito model with 70 billion parameters, reaching 84.9% accuracy. The model produced three times more CO emissions than similar sized models that generated concise answers. "Currently, we see a clear accuracy-sustainability trade-off inherent in LLM technologies," said Dauner. "None of the models that kept emissions below 500 grams of CO2 equivalent achieved higher than 80% accuracy on answering the 1,000 questions correctly." CO equivalent is the unit used to measure the climate impact of various greenhouse gases. Subject matter also resulted in significantly different levels of CO emissions. Questions that required lengthy reasoning processes, for example abstract algebra or philosophy, led to up to six times higher emissions than more straightforward subjects, like high school history. Practicing thoughtful use The researchers said they hope their work will cause people to make more informed decisions about their own AI use. "Users can significantly reduce emissions by prompting AI to generate concise answers or limiting the use of high-capacity models to tasks that genuinely require that power," Dauner pointed out. Choice of model, for instance, can make a significant difference in CO emissions. For example, having DeepSeek R1 (70 billion parameters) answer 600,000 questions would create CO emissions equal to a round-trip flight from London to New York. Meanwhile, Qwen 2.5 (72 billion parameters) can answer more than three times as many questions (about 1.9 million) with similar accuracy rates while generating the same emissions. The researchers said that their results may be impacted by the choice of hardware used in the study, an emission factor that may vary regionally depending on local energy grid mixes, and the examined models. These factors may limit the generalizability of the results. "If users know the exact CO2 cost of their AI-generated outputs, such as casually turning themselves into an action figure, they might be more selective and thoughtful about when and how they use these technologies," Dauner concluded.

Advanced AI models generate up to 50 times more CO₂ emissions than more common LLMs when answering the same questions

The processes used by advanced reasoning models generate significantly more emissions than those of conventional peers. (Image credit: Getty Images) The more accurate we try to make AI models, the bigger their carbon footprint -- with some prompts producing up to 50 times more carbon dioxide emissions than others, a new study has revealed. Reasoning models, such as Anthropic's Claude, OpenAI's o3 and DeepSeek's R1, are specialized large language models (LLMs) that dedicate more time and computing power to produce more accurate responses than their predecessors. Yet, aside from some impressive results, these models have been shown to face severe limitations in their ability to crack complex problems. Now, a team of researchers has highlighted another constraint on the models' performance -- their exorbitant carbon footprint. They published their findings June 19 in the journal Frontiers in Communication. "The environmental impact of questioning trained LLMs is strongly determined by their reasoning approach, with explicit reasoning processes significantly driving up energy consumption and carbon emissions," study first author Maximilian Dauner, a researcher at Hochschule München University of Applied Sciences in Germany, said in a statement. "We found that reasoning-enabled models produced up to 50 times more CO₂ emissions than concise response models." To answer the prompts given to them, LLMs break up language into tokens -- word chunks that are converted into a string of numbers before being fed into neural networks. These neural networks are tuned using training data that calculates the probabilities of certain patterns appearing. They then use these probabilities to generate responses. Reasoning models further attempt to boost accuracy using a process known as "chain-of-thought." This is a technique that works by breaking down one complex problem into smaller, more digestible intermediary steps that follow a logical flow, mimicking how humans might arrive at the conclusion to the same problem. Related: AI 'hallucinates' constantly, but there's a solution However, these models have significantly higher energy demands than conventional LLMs, posing a potential economic bottleneck for companies and users wishing to deploy them. Yet, despite some research into the environmental impacts of growing AI adoption more generally, comparisons between the carbon footprints of different models remain relatively rare. To examine the CO₂ emissions produced by different models, the scientists behind the new study asked 14 LLMs 1,000 questions across different topics. The different models had between 7 and 72 billion parameters. The computations were performed using a Perun framework (which analyzes LLM performance and the energy it requires) on an NVIDIA A100 GPU. The team then converted energy usage into CO₂ by assuming each kilowatt-hour of energy produces 480 grams of CO₂. Their results show that, on average, reasoning models generated 543.5 tokens per question compared to just 37.7 tokens for more concise models. These extra tokens -- amounting to more computations -- meant that the more accurate reasoning models produced more CO₂. The most accurate model was the 72 billion parameter Cogito model, which answered 84.9% of the benchmark questions correctly. Cogito released three times the CO₂ emissions of similarly sized models made to generate answers more concisely. "Currently, we see a clear accuracy-sustainability trade-off inherent in LLM technologies," said Dauner. "None of the models that kept emissions below 500 grams of CO₂ equivalent [total greenhouse gases released] achieved higher than 80% accuracy on answering the 1,000 questions correctly." But the issues go beyond accuracy. Questions that needed longer reasoning times, like in algebra or philosophy, caused emissions to spike six times higher than straightforward look-up queries. The researchers' calculations also show that the emissions depended on the models that were chosen. To answer 60,000 questions, DeepSeek's 70 billion parameter R1 model would produce the CO₂ emitted by a round-trip flight between New York and London. Alibaba Cloud's 72 billion parameter Qwen 2.5 model, however, would be able to answer these with similar accuracy rates for a third of the emissions. The study's findings aren't definitive; emissions may vary depending on the hardware used and the energy grids used to supply their power, the researchers emphasized. But they should prompt AI users to think before they deploy the technology, the researchers noted.. "If users know the exact CO₂ cost of their AI-generated outputs, such as casually turning themselves into an action figure, they might be more selective and thoughtful about when and how they use these technologies," Dauner said.

Why Some AI Models Spew 50 Times More Greenhouse Gas to Answer the Same Question

It's possible for you to make your LLM use "greener," according to new research. Like it or not, large language models have quickly become embedded into our lives. And due to their intense energy and water needs, they might also be causing us to spiral even faster into climate chaos. Some LLMs, though, might be releasing more planet-warming pollution than others, a new study finds. Queries made to some models generate up to 50 times more carbon emissions than others, according to a new study published in Frontiers in Communication. Unfortunately, and perhaps unsurprisingly, models that are more accurate tend to have the biggest energy costs. It’s hard to estimate just how bad LLMs are for the environment, but some studies have suggested that training ChatGPT used up to 30 times more energy than the average American uses in a year. What isn’t known is whether some models have steeper energy costs than their peers as they’re answering questions. Researchers from the Hochschule München University of Applied Sciences in Germany evaluated 14 LLMs ranging from 7 to 72 billion parametersâ€"the levers and dials that fine-tune a model's understanding and language generationâ€"on 1,000 benchmark questions about various subjects. LLMs convert each word or parts of words in a prompt into a string of numbers called a token. Some LLMs, particularly reasoning LLMs, also insert special “thinking tokens†into the input sequence to allow for additional internal computation and reasoning before generating output. This conversion and the subsequent computations that the LLM performs on the tokens use energy and releases CO2. The scientists compared the number of tokens generated by each of the models they tested. Reasoning models, on average, created 543.5 thinking tokens per question, whereas concise models required just 37.7 tokens per question, the study found. In the ChatGPT world, for example, GPT-3.5 is a concise model, whereas GPT-4o is a reasoning model. This reasoning process drives up energy needs, the authors found. “The environmental impact of questioning trained LLMs is strongly determined by their reasoning approach,†study author Maximilian Dauner, a researcher at Hochschule München University of Applied Sciences, said in a statement. “We found that reasoning-enabled models produced up to 50 times more CO2 emissions than concise response models.†The more accurate the models were, the more carbon emissions they produced, the study found. The reasoning model Cogito, which has 70 billion parameters, reached up to 84.9% accuracyâ€"but it also produced three times more CO2 emissions than similarly sized models that generate more concise answers. “Currently, we see a clear accuracy-sustainability trade-off inherent in LLM technologies,†said Dauner. “None of the models that kept emissions below 500 grams of CO2 equivalent achieved higher than 80% accuracy on answering the 1,000 questions correctly.†CO2 equivalent is the unit used to measure the climate impact of various greenhouse gases. Another factor was subject matter. Questions that required detailed or complex reasoning, for example abstract algebra or philosophy, led to up to six times higher emissions than more straightforward subjects, according to the study. There are some caveats, though. Emissions are very dependent on how local energy grids are structured and the models that you examine, so it’s unclear how generalizable these findings are. Still, the study authors said they hope that the work will encourage people to be “selective and thoughtful†about the LLM use. “Users can significantly reduce emissions by prompting AI to generate concise answers or limiting the use of high-capacity models to tasks that genuinely require that power,†Dauner said in a statement.

Some AI prompts could cause 50 times more CO₂ emissions than others, researchers find

No matter which questions we ask an AI, the model will come up with an answer. To produce this information -- regardless of whether the answer is correct or not -- the model uses tokens. Tokens are words or parts of words that are converted into a string of numbers that can be processed by the LLM. This conversion, as well as other computing processes, produce CO emissions. Many users, however, are unaware of the substantial carbon footprint associated with these technologies. Now, researchers in Germany measured and compared CO emissions of different, already trained, LLMs using a set of standardized questions. "The environmental impact of questioning trained LLMs is strongly determined by their reasoning approach, with explicit reasoning processes significantly driving up energy consumption and carbon emissions," said first author Maximilian Dauner, a researcher at Hochschule München University of Applied Sciences and first author of the Frontiers in Communication study. "We found that reasoning-enabled models produced up to 50 times more CO emissions than concise response models." 'Thinking' AI causes most emissions The researchers evaluated 14 LLMs ranging from seven to 72 billion parameters on 1,000 benchmark questions across diverse subjects. Parameters determine how LLMs learn and process information. Reasoning models, on average, created 543.5 "thinking" tokens per question, whereas concise models required just 37.7 tokens per question. Thinking tokens are additional tokens that reasoning LLMs generate before producing an answer. A higher token footprint always means higher CO emissions. It doesn't, however, necessarily mean the resulting answers are more correct, as elaborate detail is not always essential for correctness. The most accurate model was the reasoning-enabled Cogito model with 70 billion parameters, reaching 84.9% accuracy. The model produced three times more CO emissions than similar-sized models that generated concise answers. "Currently, we see a clear accuracy-sustainability trade-off inherent in LLM technologies," said Dauner. "None of the models that kept emissions below 500 grams of CO equivalent achieved higher than 80% accuracy on answering the 1,000 questions correctly." CO equivalent is the unit used to measure the climate impact of various greenhouse gases. Subject matter also resulted in significantly different levels of CO emissions. Questions that required lengthy reasoning processes, for example abstract algebra or philosophy, led to up to six times higher emissions than more straightforward subjects, like high school history. Practicing thoughtful use The researchers said they hope their work will cause people to make more informed decisions about their own AI use. "Users can significantly reduce emissions by prompting AI to generate concise answers or limiting the use of high-capacity models to tasks that genuinely require that power," Dauner pointed out. Choice of model, for instance, can make a significant difference in CO emissions. For example, having DeepSeek R1 (70 billion parameters) answer 600,000 questions would create CO emissions equal to a round-trip flight from London to New York. Meanwhile, Qwen 2.5 (72 billion parameters) can answer more than three times as many questions (about 1.9 million) with similar accuracy rates while generating the same emissions. The researchers said that their results may be impacted by the choice of hardware used in the study, an emission factor that may vary regionally depending on local energy grid mixes, and the examined models. These factors may limit the generalizability of the results. "If users know the exact CO cost of their AI-generated outputs, such as casually turning themselves into an action figure, they might be more selective and thoughtful about when and how they use these technologies," Dauner concludes.

AI users have to choose between accuracy or sustainability

Cheap or free access to AI models keeps improving, with Google the latest firm to make its newest models available to all users, not just paying ones. But that access comes with one cost: the environment. In a new study, German researchers tested 14 large language models (LLMs) of various sizes from leading developers such as Meta, Alibaba, and others. Each model answered 1,000 difficult academic questions spanning topics from world history to advanced mathematics. The tests ran on a powerful, energy-intensive NVIDIA A100 GPU, using a specialized framework to precisely measure electricity consumption per answer. This data was then converted into carbon dioxide equivalent emissions, providing a clear comparison of each model's environmental impact. The researchers found that many LLMs are far more powerful than needed for everyday queries. Smaller, less energy-hungry models can answer many factual questions just as well. The carbon and water footprints of a single prompt vary dramatically depending on model size and task type. Prompts requiring reasoning, which force models to "think aloud," are especially polluting because they generate many more tokens. One model, Cogito, topped the accuracy table -- answering nearly 85% of questions correctly -- but produced three times more emissions than similar-sized models, highlighting a trade-off rarely visible to AI developers or users. (Cogito did not respond to a request for comment.) "Do we really need a 400-billion parameter GPT model to answer when World War II was, for example," says Maximilian Dauner, a researcher at Hochschule München University of Applied Sciences and one of the study's authors.

Can you choose an AI model that harms the planet less?

And some chatbots are linked to more greenhouse gas emissions than others. A study published Thursday in the journal Frontiers in Communication analyzed different generative AI chatbots' capabilities and the planet-warming emissions generated from running them. Researchers found that chatbots with bigger "brains" used exponentially more energy and answered questions more accurately -- up until a point.From uninvited results at the top of your search engine queries to offering to write your emails and helping students do homework, generative artificial intelligence is quickly becoming part of daily life as tech giants race to develop the most advanced models and attract users. All those prompts come with an environmental cost: A report last year from the Energy Department found AI could help increase the portion of the nation's electricity supply consumed by data centers from 4.4% to 12% by 2028. To meet this demand, some power plants are expected to burn more coal and natural gas. And some chatbots are linked to more greenhouse gas emissions than others. A study published Thursday in the journal Frontiers in Communication analyzed different generative AI chatbots' capabilities and the planet-warming emissions generated from running them. Researchers found that chatbots with bigger "brains" used exponentially more energy and answered questions more accurately -- up until a point. "We don't always need the biggest, most heavily trained model, to answer simple questions. Smaller models are also capable of doing specific things well," said Maximilian Dauner, a doctoral student at the Munich University of Applied Sciences and lead author of the paper. "The goal should be to pick the right model for the right task." The study evaluated 14 large language models, a common form of generative AI often referred to by the acronym LLMs, by asking each a set of 500 multiple choice and 500 free response questions across five different subjects. Dauner then measured the energy used to run each model and converted the results into carbon dioxide equivalents based on global averages.In most of the models tested, questions in logic-based subjects, like abstract algebra, produced the longest answers -- which likely means they used more energy to generate compared with fact-based subjects, such as history, Dauner said. AI chatbots that show their step-by-step reasoning while responding tend to use far more energy per question than chatbots that don't. The five reasoning models tested in the study did not answer questions much more accurately than the nine other studied models. The model that emitted the most, DeepSeek-R1, offered answers of comparable accuracy to those that generated a fourth of the amount of emissions. There is key information not captured by the study, which only included open-source LLMs: Some of the most popular AI programs made by large tech corporations, such as OpenAI's ChatGPT and Google's Gemini, were not included in the results. And because the paper converted the measured energy to emissions based on a global CO2 average, it only offered an estimate; it did not indicate the actual emissions generated by using these models, which can vary hugely depending on which country the data center running it is in. "Some regions are going to be powered by electricity from renewable sources, and some are going to be primarily running on fossil fuels," said Jesse Dodge, a senior research scientist at the Allen Institute for AI who was not affiliated with the new research. In 2022, Dodge led a study comparing the difference in greenhouse gas emissions generated by training a LLM in 16 different regions of the world. Depending on the time of year, some of the most emitting areas, like the central United States, had roughly three times the carbon intensity of the least emitting ones, such as Norway. But even with this limitation, the new study fills a gap in research on the trade-off between energy cost and model accuracy, Dodge said. "Everyone knows that as you increase model size, typically models become more capable, use more electricity and have more emissions," he said. Reasoning models, which have been increasingly trendy, are likely further bumping up energy costs, because of their longer answers. "For specific subjects an LLM needs to use more words to get to a more accurate response," Dauner said. "Longer answers and those that use a reasoning process generate more emissions." Sasha Luccioni, the AI and climate lead at Hugging Face, an AI company, said that subject matter is less important than output length, which is determined by how the model was trained. She also emphasized that the study's sample size is too small to create a complete picture of emissions from AI. "What's relevant here is not the fact that it's math and philosophy, it's the length of the input and the output," she said. Last year, Luccioni published a study that compared 88 LLMs and also found that larger models generally had higher emissions. Her results also indicated that AI text generation -- which is what chatbots do -- used 10 times as much energy compared with simple classification tasks like sorting emails into folders. Luccioni said that these kinds of "old school" AI tools, including classic search engine functions, have been overlooked as generative models have become more widespread. Most of the time, she said, the average person doesn't need to use an LLM at all. Dodge added that people looking for facts are better off just using a search engine, since generative AI can "hallucinate" false information. "We're reinventing the wheel," Luccioni said. People don't need to use generative AI as a calculator, she said. "Use a calculator as a calculator." This article originally appeared in The New York Times.

AI chatbots using reason emit more carbon than those responding concisely, study finds

A study has found that chat-based generative AI emits significantly more carbon when handling complex prompts. Reasoning-enabled models produced up to 50 times more emissions than concise ones. While these models are more accurate, researchers warn of a trade-off between accuracy and sustainability, urging optimisation for environmentally conscious AI development.A study found that carbon emissions from chat-based generative AI can be six times higher when responding to complex prompts, like abstract algebra or philosophy, compared to simpler prompts, such as high school history. "The environmental impact of questioning trained (large-language models) is strongly determined by their reasoning approach, with explicit reasoning processes significantly driving up energy consumption and carbon emissions," first author Maximilian Dauner, a researcher at Hochschule Munchen University of Applied Sciences, Germany, said. "We found that reasoning-enabled models produced up to 50 times more (carbon dioxide) emissions than concise response models," Dauner added. The study, published in the journal Frontiers in Communication, evaluated how 14 large-language models (which power chatbots), including DeepSeek and Cogito, process information before responding to 1,000 benchmark questions -- 500 multiple-choice and 500 subjective. Each model responded to 100 questions on each of the five subjects chosen for the analysis -- philosophy, high school world history, international law, abstract algebra, and high school mathematics. "Zero-token reasoning traces appear when no intermediate text is needed (e.g. Cogito 70B reasoning on certain history items), whereas the maximum reasoning burden (6.716 tokens) is observed for the Deepseek R1 7B model on an abstract algebra prompt," the authors wrote. Tokens are virtual objects created by conversational AI when processing a user's prompt in natural language. More tokens lead to increased carbon dioxide emissions. Chatbots equipped with an ability to reason, or 'reasoning models', produced 543.5 'thinking' tokens per question, whereas concise models -- producing one-word answers -- required just 37.7 tokens per question, the researchers found. Thinking tokens are additional ones that reasoning models generate before producing an answer, they explained. However, more thinking tokens do not necessarily guarantee correct responses, as the team said, elaborate detail is not always essential for correctness. Dauner said, "None of the models that kept emissions below 500 grams of CO₂ equivalent achieved higher than 80 per cent accuracy on answering the 1,000 questions correctly." "Currently, we see a clear accuracy-sustainability trade-off inherent in (large-language model) technologies," the author added. The most accurate performance was seen in the reasoning model Cogito, with a nearly 85 per cent accuracy in responses, whilst producing three times more carbon dioxide emissions than similar-sized models generating concise answers. "In conclusion, while larger and reasoning-enhanced models significantly outperform smaller counterparts in terms of accuracy, this improvement comes with steep increases in emissions and computational demand," the authors wrote. "Optimising reasoning efficiency and response brevity, particularly for challenging subjects like abstract algebra, is crucial for advancing more sustainable and environmentally conscious artificial intelligence technologies," they wrote.