Claude
Identity card
The company
- Name
- Anthropic ↗
- Founded
- 2021
- Head office
- San Francisco, California, United States
- Chief executive
- Dario Amodei
The product
- Flagship model
- Opus 5.5
- Subscription
- €21.60 per month · Pro
- API
- 4 $ / 20 $ input / output, per million tokens
The data
- Law
- United States
Usage
- Web traffic, Sept. 2026
- 9.6 %
Key points
Claude is the assistant for in-depth work, designed for long-form writing, document analysis and code; its best model is the most consistent of the 5 public performance rankings: 1st at compar:IA, 2nd in the 4 others among the 10. It does not create images, its usage limits weigh on heavy users, and Anthropic publishes no energy figure.
- Long-form writing
- Document analysis
- Code
- Create images
- Intensive use at a low price
Source: Similarweb, share of visits to generative AI sites, worldwide; records published by Similarweb and repeated by the press.
Yes if you write, analyse documents or code. Look elsewhere to create images (ChatGPT, Gemini) or if your data must stay in Europe (Vibe).
Strengths
- The most consistent of the performance rankingsAmong the 10, its best model is 1st in the compar:IA arena of the French Ministry of Culture, where internet users vote blind, and 2nd in the 4 other public rankings: behind Gemini at LMArena and Artificial Analysis, behind ChatGPT at Epoch AI and ARC-AGI-2.Source: compar:IA, French Ministry of Culture, 4 November 2026 ↗ ; LMArena, text ranking, 28 October 2026 ↗
- 1st in the Future of Life Institute safety indexIn summer 2026, the Future of Life Institute ranks Anthropic 1st of the companies it evaluates, with a C+ grade, ahead of OpenAI and Google DeepMind.Source: Future of Life Institute, AI Safety Index, July 2026 ↗
- 1st in the ADL antisemitism testClaude scores 80 out of 100 in the Anti-Defamation League test, which measures the detection of antisemitic content and the response to it: the best score of the 6 models tested.Source: Anti-Defamation League, January 2026 ↗
Weak points
- Tight usage limitsSince late August 2025, weekly caps have been added to the 5-hour limits, to curb subscribers who run code agents continuously.Source: TechCrunch, 28 July 2025 ↗
- No energy figure publishedAnthropic publishes no figure for energy, water or emissions; its stock market listing prospectus states that some planned data centres will be powered by on-site gas.Source: Anthropic prospectus (SEC), 26 October 2026 ↗
- A loss-making companyIn the 1st half of 2026, Anthropic reports revenue of 16.3 billion dollars and an operating loss of 693 million dollars.Source: Anthropic prospectus (SEC), 26 October 2026 ↗
The company Anthropic
The Anthropic entry →Company history
Anthropic was born in San Francisco in 2021. Its founders, led by Dario Amodei and his sister Daniela, come from OpenAI, which they left over disagreement about how to develop AI. They want to build powerful models with safety at the centre. The company is a public benefit corporation.
The story continues
Claude was released in March 2023. It has earned its place with companies and developers, especially with Claude Code, an agent that writes and fixes code, introduced in February 2025.
In 2026, Anthropic raises 65 billion dollars, refuses the Pentagon 2 uses of Claude (mass surveillance and autonomous weapons) and publishes its stock market listing prospectus in October.
Sources: Wikipedia, 7 November 2026 ↗ ; Anthropic, 24 February 2025 ↗ ; Orrick, Anthropic's counsel, 29 May 2026 ↗
Key dates
- 2021Founded in San Francisco by former OpenAI staff↗
- 14 March 2023Launch of Claude↗
- 24 February 2025Claude Code, a programming agent, in preview↗
- 27 February 2026Disagreement with the Pentagon over mass surveillance and autonomous weapons; the Pentagon designates it a risk↗
- 29 May 2026Funding round of 65 billion dollars↗
- 26 October 2026Stock market listing prospectus made public↗
Leadership
- Chief executive officer and co-founder
- Dario Amodei ↗
- President and co-founder
- Daniela Amodei ↗
- Chief product officer
- Ami Vora ↗
Key figures
- Valuation, May 2026
- 965 billion dollarsOrrick, Anthropic's counsel, 29 May 2026 ↗
- Revenue, 1st half of 2026
- 16.3 billion dollarsAnthropic prospectus (SEC), 26 October 2026 ↗
- Operating loss, 1st half of 2026
- 693 million dollarsAnthropic prospectus (SEC), 26 October 2026 ↗
The Claude plans
- FreeFree0 €0 $
- ProStandard21.60 €20 $
- MaxPremium108 €100 $
- UltraPremiumprice published in dollars only200 $no euro price
- Team, Standard seatTeam, per seat25.25 €25 $
- Team, Premium seatTeam, per seat126.28 €125 $
- EnterpriseEnterprise, per seatplus usage at API rates20 $no euro price
Anthropic pricing pages: incl. VAT in France, excl. tax in the United States; up to date as of 9 November 2026.
Content of the 7 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. When only the dollar price appears, the company does not publish a euro price: the amount paid in France depends on the exchange rate and VAT.
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 $- ClaudeOpus 5.520 $4 $13 times the output price of Vibe
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.
Anthropic API prices for 4 models
| Model | Input, $ per million tokens | Output, $ per million tokens |
|---|---|---|
| Fable 5.1 | 10 | 50 |
| Opus 5.5 | 4 | 20 |
| Sonnet 5.5 | 2 | 10 |
| Haiku 5.5 | 0.10 | 0.50 |
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 Claude models
- Company
- Anthropic
- Models offered
- Claude Sonnetoptional, Free and all paid plans (official page ↗)Claude Haikuoptional, Free and all paid plans (official page ↗)Claude Opuspaid plan, Pro, Max, Team, Enterprise (official page ↗)Claude Fablepaid plan, Pro (paid in usage credits), Max, Team (premium seats), Enterprise (official page ↗)Claude Fable 5.2paid plan, Pro, Team (standard seats) according to the announcement; 1st out of 71 on general intelligence (official page ↗)Claude Mythos 5.2paid plan, Cyber Verification Programme, Claude Security (Enterprise) (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 Anthropic handles conversations
- Training models on conversationsFree, Pro and Max plans: yes if the setting stays on; each user must make this choice from 2025, and the box was ticked by default according to the press. Team, Enterprise and API: never.Source: Anthropic, 28 August 2025 ↗ ; MacRumors, 28 August 2025 ↗
- Opting out of trainingTurn the setting off in the choice window, or later in the privacy settings.Source: Anthropic, 28 August 2025 ↗
- Retention period30 days if you refuse training, 5 years if you accept it. Deleted conversations are no longer used.Source: Anthropic, 28 August 2025 ↗
- Applicable lawUS law: Anthropic 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: 58 out of 100, joint 1st out of 10. Choice requested from 2025, box ticked by default according to the press; conversations kept 30 days without consent, 5 years with it; no sanction in Europe. The ethics ranking
Environment
Environment score
47 4th out of 10- Simple query 40 %50
- Advanced queries 20 %67
- Decarbonised electricity 20 %33
- Environmental approach 20 %33
Black line: average of the 10. The environmental ranking
Our view
Claude generates neither images nor video: this is what makes it frugal in use. The simple query is served by Sonnet, a medium-sized model: 0.83 Wh per typical conversation according to CLEER, against 0.50 Wh for Gemini Flash. Anthropic publishes no figure for energy, water or emissions, and its prospectus of 26/10/2026 provides for sites powered by gas. The assistant is frugal in use, but the company publishes no environmental data.
Per-question figure published by Anthropic
Nothing publishedAnthropic publishes no per-question figure, neither energy nor emissions. 3 companies out of 10 publish one: Google, Mistral AI and OpenAI.
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 Claude among the 10; a click on a criterion opens its definition and the justification for Claude's score.
5 public rankings 100 %981st 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: 3rd (rank among 419 models); Artificial Analysis: 59 (index v5.0.1); Epoch AI: 168.4 (ECI index); compar:IA: 1st (rank among 121 models); ARC-AGI-2: 93.3 (% solved).
Content produced 30 %921st out of 10
Definition. Content the assistant produces on request (sexual, misleading images, minors) and results of independent tests.
Justification. Sexual content prohibited, fiction included; no images or video; ages 18 and over; 1st in the ADL test (80/100, January 2026); misused for espionage, as revealed by Anthropic (11/2025).
Use of data 30 %58joint 1st out of 10
Definition. Use of conversations for training, retention period, storage location, sanctions by the authorities.
Justification. Choice requested from 2025, box ticked by default according to the press; conversations kept 30 days without consent, 5 years with it; no sanction in Europe.
Safety 20 %831st out of 10
Definition. Independent safety evaluations (Future of Life Institute, SaferAI, Phare) and resistance to attacks.
Justification. 1st at the Future of Life Institute (C+, summer 2026) and SaferAI (35%, July 2026); 5 first places in the Phare ranking (7/2026); pause commitment abandoned in February 2026.
Governance 20 %751st out of 10
Definition. Control of the company and deliberate choices: military uses, shareholding, commitments kept or abandoned.
Justification. Refusal of mass surveillance and autonomous weapons to the Pentagon (26/2/2026), sanctioned the next day; military contract capped at $200m (7/2025).
Simple query 40 %50joint 5th 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. Default Sonnet: 0.83 Wh per typical conversation according to CLEER (Sonnet 5), against 0.50 Wh for Gemini 3.7 Flash; Haiku, 0.08 Wh, remains available.
Advanced queries 20 %672nd out of 10
Definition. Weight of the resource-hungry uses the product promotes: video and image generation, agents, reasoning enabled by default.
Justification. No images or video; code agents sometimes run continuously (weekly limits since August 2025).
Decarbonised electricity 20 %33joint 4th 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. Mostly in the United States; the prospectus of 26/10/2026 provides for sites powered by on-site gas.
Environmental approach 20 %33joint 6th out of 10
Definition. What the company publishes and commits to: energy and emissions figures, environmental report, targets, model optimisation.
Justification. No figure for energy, water or emissions; Opus 5.5 "less costly to serve" (September 2026).
Compare
or Claude or ChatGPTClaude is better rated on performance and ethics; ChatGPT does more things, images and voice included, and has far more users.Celsius score: 74 against 62 or Claude or GeminiGemini is better rated on environment and integrated with Google; Claude is better rated on performance and ethics.Celsius score: 74 against 71 or Claude or VibeVibe keeps data in Europe and costs less; Claude is markedly more performant.Celsius score: 74 against 40Other comparisons
Claude or Copilot+16Claude or Meta AI+26Claude or Qwen+35Claude or Perplexity+38Claude or DeepSeek+39Claude or Grok+48Celsius score gap.
Ranked 1st 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: Anthropic, S-1 prospectus (SEC, 26/10/2026), 11 November 2026 ↗.
Continue in the ranking
The full ranking →CompanyAnthropic12models measuredIts assistant, its models, its governance and the electricity of its data centres.The entry →ModelAnthropicClaude Fable 5.21stout of 71 on general intelligenceThe best ranked on general intelligence of the models Claude offers.The entry →Comparison2 AI assistantsClaude or ChatGPT74 against 62Celsius score out of 100Claude is better rated on performance and ethics; ChatGPT does more things, images and voice included, and has far more users.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
