---
title: "Which AI should you choose for the smallest footprint?"
description: "The model matters more than the brand: 0.05 Wh per query for a small model, 0.3 Wh for a standard one, 2.3 Wh for a large one that thinks. Figures and tips."
canonical: https://projetcelsius.com/en/blog/chatgpt-vs-claude-vs-gemini-environmental-impact/
author: "Guillaume Pakula"
datePublished: 2026-06-08
dateModified: 2026-11-03
language: en
translationOf: https://projetcelsius.com/blog/ia-comparatif-chatgpt-claude-gemini/
---

# Which AI should you choose for the smallest footprint?

**Author**: Guillaume Pakula
**Published**: 2026-06-08
**Updated**: 2026-11-03
**Section**: Explained
**Canonical URL**: https://projetcelsius.com/en/blog/chatgpt-vs-claude-vs-gemini-environmental-impact/
**French original**: https://projetcelsius.com/blog/ia-comparatif-chatgpt-claude-gemini/
**Source**: Projet Celsius (https://projetcelsius.com/en/), a climate strategy consultancy of engineers and PhDs based in Paris and Marseille.

> **The type of model matters far more than the brand**: a query costs about **0.05 Wh** on a small model, **0.3 Wh** on an assistant's standard model and **2.3 Wh** on a large model in thinking mode, according to Projet Celsius's AI Impact Database. ChatGPT, Claude, Gemini and Mistral AI all offer these ranges, and only Google and Mistral AI publish [a figure with its method](https://arxiv.org/abs/2508.15734). Choosing the smallest model that does the job counts for more than switching assistant.

## Key takeaways

- **The model matters more than the brand**: a query uses 0.05 to 1.2 Wh depending on the model.
- **Thinking mode doubles energy use** on average, and far more on coding tasks.
- **Only Google and Mistral AI publish a figure** with its method; Anthropic publishes nothing.
- **A 5-second video clip** uses as much energy as 330 queries to an AI.

In France, **48% of people aged 12 and over** used generative AI in 2025, up from **20% in 2023**, according to the [Digital Barometer](https://www.arcep.fr/uploads/tx_gspublication/barometre-du-numerique-edition-2026_INFOGRAPHIE.pdf) produced by the Crédoc research centre for Arcep (the French electronic communications regulator). Each query weighs little, but their number shows up in data centres, the buildings where the servers run: their global electricity consumption is projected to rise from about **415 TWh in 2024 to nearly 950 TWh in 2030** in the base case of the [International Energy Agency (IEA)](https://www.iea.org/reports/key-questions-on-energy-and-ai).

The answer to the question in the title therefore lies **less in a logo than in a setting**. ChatGPT, Claude, Gemini and Mistral AI's assistant each offer a light model, a default model and a more powerful model, and most add a **thinking mode**. The figures in this article come from the [AI Impact Database (in French)](https://projetcelsius.com/ia/base/), Projet Celsius's database of AI impacts, in watt-hours (Wh): a [phone charge](https://projetcelsius.com/en/blog/company-carbon-footprint-main-emission-sources/) takes 19.

## Which AI uses the least energy, model by model?

For a query and an answer half a page long, a standard-sized assistant uses **about 0.3 Wh**, or 0.14 gCO2e (grams of CO2 equivalent) with the global average electricity mix. That is the value on which a Google measurement, 2 estimates and a statement from OpenAI converge, as set out in our article on [the energy used by a ChatGPT query](https://projetcelsius.com/en/blog/chatgpt-query-energy-consumption/), roughly the same as [a Google search](https://projetcelsius.com/en/blog/chatgpt-vs-google-search-energy/). The same query costs **about 0.05 Wh on a small model** and **1.2 Wh on a large model**.

### Small, standard or large: model size

A larger model performs **more computation for every word** it writes, and electricity follows computation. The 4 main assistants rank their models in the same way, from the lightest to the most powerful:

- At Anthropic, Claude comes as **Haiku**, designed for fast, high-volume tasks, then Sonnet, Opus and Fable, reserved for the most demanding reasoning.
- At Google, Gemini offers **Flash-Lite**, the fastest and cheapest, then Flash and Pro, which the app presents as the model that thinks through complex topics.
- At OpenAI, ChatGPT answers by default with a fast model, **Instant**, and can switch to its Thinking mode, which reasons before answering.
- At Mistral AI, the Le Chat assistant, renamed Vibe in 2026, relies on a range that runs from **Mistral Small**, the lightest, to Mistral Large.

**None of these providers publishes the size** of its models. Classing them as small, standard and large remains **an assumption based on their place in the range** and on their price, which partly tracks the amount of computation: according to its [public price list](https://platform.claude.com/docs/en/about-claude/models/overview), Anthropic charges developers **40 times less for Haiku than for Opus**. The values in the opening chart are those of the AI Impact Database for a typical size. At Anthropic, the choice between the 2 large models is covered in [Fable 5.2 or Opus 5.5](https://projetcelsius.com/en/blog/fable-5-2-vs-opus-5-5/).

An academic study gives an order of magnitude model by model. In 2025 its authors estimated the electricity used by a short query followed by an answer of about 200 words, based on each model's public response speed and the likely hardware in the data centres: these are estimates, not measurements by the providers. According to ["How Hungry is AI?"](https://arxiv.org/abs/2505.09598) (Jegham et al., November 2025 version):

- GPT-4.1 nano, OpenAI's smallest model in the study, uses **0.21 Wh** per query.
- GPT-4o, ChatGPT's default model until August 2025, uses **0.42 Wh**, or 0.19 gCO2e with the global average electricity mix.
- Anthropic's Claude 3.7 Sonnet uses **0.95 Wh**, almost as much as Claude 3.5 Haiku (0.98 Wh), even though Haiku was the fast model in the range at the time.
- o3, OpenAI's reasoning model, uses **1.2 Wh**, and o4-mini with high reasoning effort **3.6 Wh**.

Almost all of these 2025 models have since been replaced. The Haiku case shows that a "fast" model name does not guarantee an economical query, and that the method sees neither the actual size of the models nor the exact machines. The orders of magnitude match those of the AI Impact Database: a few tenths of a watt-hour for a standard model, more than 1 Wh for a large model or a model that thinks.

### Thinking mode: energy use doubled on average

Thinking mode has the model write **a draft you do not see** before its answer. Across 22 open models measured by PEReN, the French digital regulation expertise centre, it increases energy use **by 92% on average**, and by 849% on average on programming exercises, according to [Arcep's May 2026 report](https://www.arcep.fr/uploads/tx_gspublication/rapport-IA-et-environnement_210526.pdf). A large model in thinking mode therefore uses about 2.3 Wh for a query, **46 times more than a small model** answering directly.

> Within the same assistant, the choice of model and of thinking mode changes the footprint of a query more than the choice of assistant.

## What do OpenAI, Anthropic, Google and Mistral publish?

**Only 2 providers** have published a figure together with its method, and **no figure compares directly** with the others: they count neither the same thing nor the same query.

### A measurement, a sentence, an assessment and a silence

Google measured [the median query to Gemini (in French)](https://projetcelsius.com/ia/chiffres/electricite-requete-gemini/) in production: **0.24 Wh and 0.26 mL of water**, the equivalent of 5 drops, according to its [August 2025 paper](https://arxiv.org/abs/2508.15734). The same paper states that the energy of this query was **cut 33-fold in one year**. It is the only production measurement published by a provider, and no third party has audited it.

OpenAI sticks to a single sentence from its chief executive, Sam Altman: an average query would use **about 0.34 Wh**, according to [his June 2025 blog post](https://blog.samaltman.com/the-gentle-singularity), with no method or scope. Anthropic publishes **no figure** for Claude. In July 2025, Mistral AI published the first [life cycle assessment](https://projetcelsius.com/en/blog/lca-vs-carbon-footprint/) of a large model, carried out with ADEME (the French Agency for Ecological Transition): **1.14 gCO2e and 45 mL of water** for one answer from its assistant, [server manufacturing included](https://mistral.ai/news/our-contribution-to-a-global-environmental-standard-for-ai).

### Which AI uses the least water?

Google's 0.26 mL counts only **the water evaporated to cool** its data centres. Mistral AI's 45 mL counts the water used across **the whole life cycle**, including chip manufacturing and electricity generation. **The gap comes from the scope**, and does not say which of the 2 assistants uses less water. Our article on [the water AI uses](https://projetcelsius.com/en/blog/ai-water-footprint/) details these 3 kinds of water, and the one on [the stages at which AI harms the environment](https://projetcelsius.com/en/blog/why-ai-is-bad-for-the-environment/) the weight of manufacturing.

## Which habits reduce your AI footprint?

Model size is not the only lever. **The nature of the task** makes electricity use vary far more: from 0.13 Wh for the summary generated at the top of a search to **670 Wh for [one hour of a coding agent (in French)](https://projetcelsius.com/ia/base/taches/heure-d-agent-de-code/)**, a program that chains dozens of steps on its own.

5 habits reduce the footprint without giving up the tool, from the simplest to the most effective depending on use:

- Start with **your assistant's lightest model**, and move up only if the answer is not good enough: for a query, the gap runs from 0.05 to 1.2 Wh.
- Leave **thinking mode off** for simple questions, a fact search or an email: it doubles energy use on average, for a gain in quality that varies from task to task according to Arcep.
- **Attach a document** only when the answer depends on it: querying a 30-page PDF costs about 4.7 Wh, as much as 16 queries, because the model reads the whole text.
- Keep **images and video** for real needs: a generated image takes about 3.2 Wh, and a 5-second clip nearly 100 Wh, as much as 330 queries.
- Open **a new conversation** when you change subject: with each message, the model rereads the earlier exchange, and 10 exchanges already use about 3.5 Wh.

Our article on [10 habits ranked by effectiveness](https://projetcelsius.com/en/blog/reduce-ai-environmental-impact/) quantifies them over a year of use, and those on [the footprint of an AI image](https://projetcelsius.com/en/blog/ai-image-carbon-footprint/) and [of an AI-generated video](https://projetcelsius.com/en/blog/ai-video-environmental-impact/) cover the heaviest uses.

## Ecosia, Mistral: is there such a thing as green AI?

**No AI is green in itself.** Some combine more favourable conditions: a small model, low-carbon electricity, published figures. Our comparison of [AI tools marketed as eco-friendly](https://projetcelsius.com/en/blog/most-eco-friendly-ai/) covers Ecosia AI, Infomaniak's Euria, GreenPT and Vibe.

### Ecosia AI, a small Mistral AI model

The Ecosia search engine runs its AI on **Mistral AI models**, including **Mistral Small 3.2** for the summaries at the top of results, according to its [help centre](https://support.ecosia.org/article/1006-ai-search?hl=en-US). In a [blog post of 2 July 2026](https://blog.ecosia.org/ai-free-search/), it claims to **produce more clean energy than its AI uses**, without publishing a figure. Its AI features **can be switched off in the settings**, and it estimates the impact of each request with the EcoLogits method, which leaves out training, networks and the user's device.

### Mistral AI and the country where the servers run

Mistral AI remains **the only provider to have published a life cycle assessment**, reviewed by 2 engineering consultancies. It does not say that its assistant emits less than ChatGPT or Gemini: its 1.14 gCO2e include server manufacturing, which the other figures leave out.

The servers' location matters too. Producing 1 kWh emits **41 gCO2e in France**, against 384 in the United States and 458 on the global average, according to the Ember data used by the AI Impact Database: the same query therefore emits **about 11 times less** in France. Yet **no major assistant states** where the answer to a given query is computed.

## Does AI weigh heavily in a company's footprint?

In a company, the same reasoning applies at a larger scale, and it is **intensive uses** that weigh. One hour of a coding agent uses about 670 Wh, or **310 gCO2e** with global electricity, the equivalent of 35 phone charges. Setting **a light model as the default** in internal tools, then reserving large models and thinking mode for the teams that need them, is the first lever, and it costs nothing.

Our article on [AI in a company's Bilan Carbone®](https://projetcelsius.com/en/blog/ai-carbon-footprint-in-your-company/), the French carbon accounting method, explains how to estimate this item and what it represents. The [AI carbon footprint calculator](https://projetcelsius.com/en/tools/ai-carbon-footprint-calculator/) gives an order of magnitude, and the method for [including AI in your Bilan Carbone®](https://projetcelsius.com/en/blog/ai-emissions-carbon-accounting-scope-3/) is covered separately, as is [a company's digital footprint](https://projetcelsius.com/en/blog/corporate-digital-carbon-footprint/).

For us at Projet Celsius, the useful question in front of an assistant is "what is the smallest model that does this job properly?". The providers **do not publish enough to measure each brand on the same basis**: our [environmental ranking of AI assistants (in French)](https://projetcelsius.com/ia/palmares/environnement/) compares them on 4 criteria, from figures of different kinds. Each of them does, however, let you choose the model size and thinking mode, and these 2 settings make the electricity used by a query vary **from 1 to nearly 50**. We recommend that individuals **start with the lightest model**, and that companies make it **the default setting** in their tools.

## Key takeaways

- The type of model matters more than the brand: a query costs about 0.05 Wh on a small model, 0.3 Wh on a standard model and 1.2 Wh on a large one.
- Thinking mode increases energy use by 92% on average according to PEReN's measurements, and by 849% on average on programming exercises.
- Google publishes a measurement (0.24 Wh), OpenAI a sentence (0.34 Wh), Mistral AI a life cycle assessment (1.14 gCO2e), and Anthropic nothing for Claude.
- The task weighs more than the assistant: a 5-second video clip uses as much as 330 queries, one hour of a coding agent as much as 2,200.
- No AI is green in itself: a small model, without thinking mode, on servers powered by low-carbon electricity remains the most economical choice.

To place these figures among AI's wider impacts, see [all the environmental impacts of AI](https://projetcelsius.com/en/blog/environmental-impact-of-ai/), [AI's electricity consumption worldwide](https://projetcelsius.com/en/blog/ai-electricity-consumption/) and [should you stop using AI?](https://projetcelsius.com/en/blog/should-you-boycott-ai/)

## Frequently asked questions

### Which AI has the lowest environmental impact?

No brand does across the board: the model is what counts. A query costs about 0.05 Wh on a small model, 0.3 Wh on a standard model and 1.2 Wh on a large one, according to Projet Celsius's AI Impact Database. The most economical choice is your assistant's smallest model (Haiku for Claude, Flash-Lite for Gemini), without thinking mode, when it is enough for the task.

### Claude, ChatGPT or Gemini: which uses the least energy?

They cannot be ranked: Google publishes a measurement of 0.24 Wh per Gemini query (August 2025), OpenAI a statement of 0.34 Wh for ChatGPT (June 2025), and Anthropic no figure for Claude. All 3 assistants offer a light model, a standard model and a powerful model: the gap between these models, from 1 to about 25 for a query, far exceeds the gap between the brands.

### Which AI uses the least water?

Google measured 0.26 mL of water for a median Gemini query, or 5 drops, counting only the water evaporated to cool its data centres. Mistral AI publishes 45 mL per answer, but over the whole life cycle, including server manufacturing and electricity generation. The 2 figures do not share the same scope and cannot be used to rank the assistants.

### Does thinking mode use more energy?

Yes. Thinking mode has the model write an invisible draft before its answer. Across 22 open models measured by PEReN for Arcep (May 2026 report), it increases energy use by 92% on average, and by 849% on average on programming exercises. It is therefore best kept for difficult questions, where it genuinely improves the answer.

### Is Ecosia an eco-friendly AI?

Ecosia AI uses Mistral AI models, including Mistral Small 3.2 for the summaries at the top of results, and claims to produce more clean energy than its AI uses (blog post of 2 July 2026), without a published figure. Its AI features can be switched off in the settings. A small model is a step in the right direction, but no AI is green in itself.

### How much energy does an AI query use?

About 0.3 Wh for a query and a half-page answer on a standard-sized assistant, or 0.14 gCO2e with the global average electricity mix (458 gCO2e per kWh, Ember). That is the value on which Google's measurement (0.24 Wh for Gemini), OpenAI's statement (0.34 Wh) and 2 independent estimates converge. A phone charge is worth about 63 such queries.

## Sources

**Sources**: Projet Celsius, AI Impact Database, engine values as at 3 November 2026, coding agent hour (IA-0048) at version 0.4 of 12 November 2026 (tasks IA-0042 to IA-0048, IA-0201 to IA-0216, global kilowatt-hour at 458 gCO2e based on Ember); [Google, Measuring the environmental impact of delivering AI at Google scale, arXiv 2508.15734, August 2025](https://arxiv.org/abs/2508.15734); [Sam Altman, The Gentle Singularity, June 2025](https://blog.samaltman.com/the-gentle-singularity); [Mistral AI, Our contribution to a global environmental standard for AI, 22 July 2025](https://mistral.ai/news/our-contribution-to-a-global-environmental-standard-for-ai); [Arcep, Intelligence artificielle et environnement, May 2026 (PEReN study, in French)](https://www.arcep.fr/uploads/tx_gspublication/rapport-IA-et-environnement_210526.pdf); [Crédoc for Arcep, Baromètre du numérique, 2026 edition (in French)](https://www.arcep.fr/uploads/tx_gspublication/barometre-du-numerique-edition-2026_INFOGRAPHIE.pdf); [IEA, Key Questions on Energy and AI, April 2026](https://www.iea.org/reports/key-questions-on-energy-and-ai); [Anthropic, Models overview, accessed 3 November 2026](https://platform.claude.com/docs/en/about-claude/models/overview); [Google AI Studio, list of Gemini models, accessed 3 November 2026](https://aistudio.google.com/docs/models); [Ecosia, AI-free search, 2 July 2026](https://blog.ecosia.org/ai-free-search/); [Ecosia, help centre, AI search, accessed 3 November 2026](https://support.ecosia.org/article/1006-ai-search?hl=en-US); [EPA, Greenhouse Gas Equivalencies Calculator (phone charge, 0.019 kWh)](https://www.epa.gov/energy/greenhouse-gas-equivalencies-calculator-calculations-and-references); [Jegham et al., How Hungry is AI? Benchmarking Energy, Water, and Carbon Footprint of LLM Inference, arXiv 2505.09598, version 6, November 2025 (estimate based on public throughput measured by Artificial Analysis)](https://arxiv.org/abs/2505.09598).
