Gemini
Identity card
The company
- Name
- Google ↗
- Founded
- 1998
- Head office
- Mountain View, California, United States
- Chief executive
- Sundar Pichai
The product
- Flagship model
- Gemini 4 Argon
- Subscription
- €21.99 per month · Google AI Pro
- API
- 2 $ / 10 $ input / output, per million tokens
The data
- Law
- United States
Usage
- Web traffic, Sept. 2026
- 25.0 %
Key points
Gemini is built into Search, Android and Gmail, and relies on a model that tops several rankings. It keeps your conversations for 18 months by default and too easily produces misleading images.
- Android and Gmail users
- Search and synthesis
- Images and video
- Keeping your exchanges private
- Teenagers
Source: Similarweb, share of visits to generative AI sites, worldwide; records published by Similarweb and repeated by the press.
Yes if you already use Gmail, Android or Google Search, or if you want the best balance between performance and frugality. Look elsewhere if the privacy of your exchanges matters most (Claude, Vibe).
Strengths
- 1st in 2 major rankingsIts best model is 1st out of 419 in the LMArena text ranking and 1st in the Artificial Analysis index.Source: LMArena, text ranking, 28 October 2026 ↗ ; Artificial Analysis, Intelligence Index, 4 November 2026 ↗
- The most complete measurement of the energy of a question3 companies out of 10 publish a per-question figure; Google is the only one to publish the details of its measurement method: 0.24 Wh and 0.26 mL of water for the median question (August 2025), 33 times less energy than a year earlier.Source: Google Cloud, 21 August 2025 ↗
- 1 billion usersThe Gemini app has passed one billion users; it is the product that has grown fastest in Google's history, according to Sundar Pichai.Source: Google, 3rd quarter 2026 results, 28 October 2026 ↗
Weak points
- Conversations kept for 18 monthsBy default, activity is kept for 18 months and may be used to train the models; conversations reviewed by humans are kept for up to 3 years, even after deletion.Source: Google, Gemini privacy hub, 7 November 2026 ↗
- Misleading images easy to obtainNewsGuard obtained images illustrating 30 false claims out of 30 with Nano Banana Pro, Google's image generator.Source: NewsGuard, 3 December 2025 ↗
- High risk for teenagersCommon Sense Media rates the versions for children and teenagers "high risk", for lack of real differences in responses by age.Source: TechCrunch, 5 September 2025 ↗
The company Google
The Google entry →Company history
Gemini is Google's assistant, founded in 1998 by Larry Page and Sergey Brin in California. Outpaced by ChatGPT in late 2022, Google launches Bard in March 2023, then merges its 2 research labs, Google Brain and DeepMind, under the leadership of Demis Hassabis.
The story continues
The Gemini model was released in December 2023 and gave its name to the assistant in February 2024. Google installs it everywhere: Search, Android, Gmail, Docs.
In 2026, the app passes one billion users, and Gemini 4 Argon arrives for subscribers in October.
Sources: Google, 20 April 2023 ↗ ; Google, 8 February 2024 ↗ ; Google, 3rd quarter 2026 results, 28 October 2026 ↗ ; Wikipedia, 7 November 2026 ↗
Key dates
- 21 March 2023Bard opened to the public in the United States and the United Kingdom↗
- 20 April 2023Birth of Google DeepMind↗
- 6 December 2023Presentation of the Gemini model↗
- 8 February 2024Bard becomes Gemini↗
- 21 August 2025Measurement of the energy of a question, published with its method↗
- 28 October 20261 billion app users↗
Leadership
- Chief executive officer of Google and Alphabet
- Sundar Pichai ↗
- Chief executive officer of Google DeepMind
- Demis Hassabis ↗
Key figures
- App users, October 2026
- 1 billionGoogle, 3rd quarter 2026 results, 28 October 2026 ↗
- Alphabet revenue, 2025
- 402.8 billion dollarsAlphabet, 2025 results, 4 February 2026 ↗
The Gemini plans
- FreeFree0 €0 $
- Google AI PlusStandard4.99 €4.99 $
- Google AI ProStandard21.99 €19.99 $
- Google AI UltraPremium€99.99 to €219.99$99.99 to $199.99
Google pricing pages: incl. VAT in France, excl. tax in the United States; up to date as of 9 November 2026.
Content of the 4 plans, prices in France and the United States
Prices in France and the United States. In the United States, prices are shown in dollars and excluding tax; in France, they include 20% VAT. So $20 does not equal €20: the French price is often slightly higher.
VibeMistral Large1.50 $0.50 $
PerplexitySonar2.50 $0.25 $
DeepSeekV4-Pro3.96 $1.32 $
Meta AIMuse Spark 1.14.25 $1.25 $
QwenQwen3.8-Max6 $2 $
GrokGrok 4.76 $2 $- GeminiGemini 4 Argon10 $2 $6.7 times the output price of Vibe
ClaudeOpus 5.520 $4 $
ChatGPTGPT-6 Astra50 $10 $
Copilot: no API of its own. Flagship model of each company, standard rate, up to date as of 9 November 2026.
Prices up to date as of 9 November 2026, read on the companies' pages, or from 2 concordant sources; the date of each record is in the page's price file Methodology.
The Gemini models
- Company
- Google
- Models offered
- Gemini Flashoptional, Free (official page ↗)Gemini Flashpaid plan, Google AI Plus, AI Pro, AI Ultra (official page ↗)Gemini 3.6 Flashoptional, unspecified; 41st out of 71 on general intelligence (official page ↗)Gemini 4 Argonpaid plan, Google AI Pro, AI Ultra; 7th out of 71 on general intelligence (official page ↗)Gemini Propaid plan, Google AI Pro, AI Ultra (official page ↗)
Links read on 10/11/2026 on the company's pages, press as a fallback. Rank: Epoch AI ECI index (CC BY 4.0). Performance of AI models
Data management
How Google handles conversations
- Training models on conversationsYes, as long as the "Keep activity" setting is on. Google Workspace business accounts are not used for training without the company's consent.Source: Google, Gemini privacy hub, 7 November 2026 ↗
- Opting out of trainingTurn off "Keep activity" in the Gemini settings; conversations are then kept for 72 hours, to respond and protect the service.Source: Google, Gemini privacy hub, 7 November 2026 ↗
- Retention period18 months by default, adjustable to 3 or 36 months; up to 3 years for conversations reviewed by humans.Source: Google, Gemini privacy hub, 7 November 2026 ↗
- Applicable lawUS law: Google is a US company, subject in particular to the Cloud Act, even when the servers are in Europe.Source: Guide d'usage de l'IA pour les agents publics de l'État (DINUM), June 2026 ↗
Score on this criterion: 33 out of 100, joint 4th out of 10. Activity kept 18 months by default, 3 years if reviewed by humans; Irish authority investigation opened in 2024. The ethics ranking
Environment
Environment score
73 1st out of 10- Simple query 40 %83
- Advanced queries 20 %50
- Decarbonised electricity 20 %67
- Environmental approach 20 %83
Black line: average of the 10. The environmental ranking
Our view
Of the 3 companies that publish a per-query figure, Google is the only one to publish the details of its measurement method: 0.24 Wh and 0.26 mL of water for the median text query of the Gemini app, measured in May 2025; Google does not say which model serves it. It announces 33 times less energy per query in one year and 65% carbon-free electricity hour by hour. Heavy uses are growing: video, Deep Research, 5 billion images in one month, and group emissions up 81% since 2019.
Per-question figure published by Google
0.24 Whper question, according to Google: measurement published in August 2025, median question.
Company figures cannot be compared with each other: scopes and methods differ. The "Simple query" criterion of the environmental ranking combines them with external estimates, for lack of a common basis: the methodology sets out the limits.
- Gemini0.24 Wh
Copilotnothing published- Vibe1.14 gCO2e
- Claudenothing published
- ChatGPT0.34 Wh
- Qwennothing published
- Meta AInothing published
- Perplexitynothing published
- DeepSeeknothing published
- Groknothing published
On the Éclairages blog: What is the electricity consumption of AI? ↗ ; Which AIs pollute the most? ↗
Scores and criteria
The ranking →The Celsius score is the average of 3 scores: performance, ethics and environment. Each combines several criteria. Each row places Gemini among the 10; a click on a criterion opens its definition and the justification for Gemini's score.
5 public rankings 100 %942nd out of 10
Definition. Synthesis of 5 public performance rankings, where models are ranked by tests or by votes from internet users: each ranking gives 100 to the best of the 10 and 0 to the last.
Justification. LMArena: 1st (rank among 419 models); Artificial Analysis: 60 (index v5.0.1); Epoch AI: 166.1 (ECI index); compar:IA: 5th (rank among 121 models); ARC-AGI-2: 91.3 (% solved).
Content produced 30 %58joint 4th out of 10
Definition. Content the assistant produces on request (sexual, misleading images, minors) and results of independent tests.
Justification. Strict rules, generation of people switched off in 2024; "high risk" for teenagers (Common Sense Media, 9/2025); 30 false images out of 30 obtained (NewsGuard).
Use of data 30 %33joint 4th out of 10
Definition. Use of conversations for training, retention period, storage location, sanctions by the authorities.
Justification. Activity kept 18 months by default, 3 years if reviewed by humans; Irish authority investigation opened in 2024.
Safety 20 %58joint 3rd out of 10
Definition. Independent safety evaluations (Future of Life Institute, SaferAI, Phare) and resistance to attacks.
Justification. Grade C at the Future of Life Institute, 20% at SaferAI; Gemini 2.5 Pro safety card published 22 days after the model (2025).
Governance 20 %33joint 5th out of 10
Definition. Control of the company and deliberate choices: military uses, shareholding, commitments kept or abandoned.
Justification. Pledge not to build AI for weapons and surveillance withdrawn (2/2025); military contract capped at $200m.
Simple query 40 %831st out of 10
Definition. Consumption of a typical text query, on the default model served: figure published by the company, third-party estimate, failing that the size of the model.
Justification. 0.24 Wh and 0.26 mL of water measured by Google for the median text query of the Gemini app (May 2025), without naming the model.
Advanced queries 20 %505th out of 10
Definition. Weight of the resource-hungry uses the product promotes: video and image generation, agents, reasoning enabled by default.
Justification. Video, Deep Research and images (Nano Banana, 5 billion in one month); Google itself classes video as "high" cost.
Decarbonised electricity 20 %672nd out of 10
Definition. Share of low-carbon electricity in the data centres and efforts to increase it: local grid, on-site gas, renewable or nuclear electricity contracts.
Justification. 65% carbon-free electricity hour by hour (2025), among the most efficient data centres.
Environmental approach 20 %831st out of 10
Definition. What the company publishes and commits to: energy and emissions figures, environmental report, targets, model optimisation.
Justification. 33 times less energy per question in one year, in-house TPU chips, SBTi-validated target; emissions up 81% since 2019.
Compare
or Gemini or ClaudeGemini is better rated on environment and integrated with Google's tools; Claude is better rated on performance and ethics.Celsius score: 71 against 74 or Gemini or ChatGPTThe 2 mainstream assistants: Gemini publishes its energy with its method, ChatGPT remains the most used.Celsius score: 71 against 62 or Gemini or CopilotGoogle against Microsoft: each embeds itself in its own office suite.Celsius score: 71 against 58Other comparisons
Gemini or Meta AI+23Gemini or Vibe+31Gemini or Qwen+32Gemini or Perplexity+35Gemini or DeepSeek+36Gemini or Grok+45Celsius score gap.
Ranked 2nd out of 10 in the assistant ranking, scores as of 11 November 2026. Facts and figures in the entry recorded from 7 to 9/11/2026, each with its source; foundation and head office: Google, "Our story", 9 November 2026 ↗.
Continue in the ranking
The full ranking →CompanyGoogle11models measured, of which 8 with an entryIts assistant, its models, its governance and the electricity of its data centres.The entry →ModelGoogleGemini 4 Argon7thout of 71 on general intelligenceThe best ranked on general intelligence of the models Gemini offers.The entry →Comparison2 AI assistantsGemini or Claude71 against 74Celsius score out of 100Gemini is better rated on environment and integrated with Google's tools; Claude is better rated on performance and ethics.The comparison →The hub newsletter
This month's ranking, news on the impact of AI and one figure from the AI Impact Database. One newsletter a month at most, free of charge. Past issues
