Heartipedia

Claude (language model)

Claude is an AI assistant and the family of large language models behind it, made by Anthropic and used in the claude.ai web, desktop and mobile apps, in the coding agent Claude Code and through the Claude API.[^1][^17][^32] First released on Mar 14, 2023, it is free to start, with paid plans from $20 a month.[^3][^2] As of Nov 7, 2026 the current models are Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5.[^1]

A tour of the Claude app Anthropic's own walkthrough of claude.ai: the sidebar for new and earlier chats, projects that keep conversations with shared context and instructions, and artifacts that turn an idea into an app, tool or document to share.

Developer
Anthropic, founded in 2021[24]
Type
Large language models and an AI assistant[1]
First release
Mar 14, 2023 (Claude and Claude Instant)[3]
Current models
Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5[1]
Context window
1 million tokens on every current model[1]
Where to use it
claude.ai on the web, desktop and mobile; Claude Code; the Claude API; Amazon Bedrock, Google Cloud and Microsoft Foundry[17][32][1]
App plans
Free; Pro $20 a month; Max $100; Ultra $200; Team from $25 a seat[2]
API prices
From $0.10 (Haiku 5.5) to $10 (Fable 5.1) per million input tokens[1]

Overview

Claude is the name of both an AI assistant and the family of large language models that powers it, made by Anthropic, a company that Dario Amodei co-founded in 2021 after leaving OpenAI with a small group of researchers.[24][1] People use it in three main ways: the claude.ai website with its desktop and mobile apps, the coding agent Claude Code, and the Claude API that developers build into their own software.[17][32][1] The models are also sold through Amazon Bedrock, Google Cloud and Microsoft Foundry.[1] Every current Claude model reads text and images and writes text, and can call tools such as web search and code execution.[1] Claude does not generate photos or illustrations, a capability that OpenAI's ChatGPT offers.[14][29] Among commercial systems of artificial intelligence, Claude competes most directly with ChatGPT and Google's Gemini, and on Nov 6, 2026 a Claude model led the public WebDev Arena leaderboard for building web apps.[29][28][27]

A tour of the Claude app Anthropic's own walkthrough of claude.ai: the sidebar for new and earlier chats, projects that keep conversations with shared context and instructions, and artifacts that turn an idea into an app, tool or document to share.

Early versions (2023 to 2024)

Anthropic released the first Claude on Mar 14, 2023, together with Claude Instant, a lighter, faster and cheaper model, through a chat interface and an API, with partners including Notion, Quora's Poe app and DuckDuckGo.[3] Claude 2 followed on Jul 11, 2023, with a public beta of the claude.ai website in the US and UK and prompts of up to 100,000 tokens; Anthropic reported a score of 76.5% on the bar exam's multiple-choice section.[4] The first paid plan, Claude Pro, launched on Sep 7, 2023 at $20 a month in the US and 18 pounds in the UK.[5] On Mar 4, 2024 the Claude 3 family introduced the three names still in use: Haiku, the fastest and smallest model, Sonnet in the middle, and Opus, the most capable.[6] Claude 3 could read images such as photos, charts and diagrams and took 200,000 tokens of context; Sonnet powered the free site, while Opus needed Pro.[6] On Oct 22, 2024 an upgraded Claude 3.5 Sonnet became, in Anthropic's words, the first frontier model to offer computer use in public beta, operating a computer through screenshots, cursor moves, clicks and typing.[7] Anthropic called the feature "experimental" and admitted it was "at times cumbersome and error-prone".[7]

Computer use fills in a form In Anthropic's October 2024 demonstration, Claude takes screenshots, notices that a company is missing from a spreadsheet, looks it up in a CRM, scrolls for the details and copies them into a vendor form before submitting it, with no help from the person watching.

Claude Code and reasoning models (2025)

Claude 3.7 Sonnet, released on Feb 24, 2025, was presented by Anthropic as the first hybrid reasoning model on the market: it could answer at once or reason step by step, showing its thinking, before replying.[8] The same announcement introduced Claude Code, a command-line tool for programmers, as a limited research preview that could search and read code, edit files, and write and run tests from the terminal.[8] Claude Opus 4 and Sonnet 4 followed on May 22, 2025; Anthropic called Opus 4 the best coding model in the world, made Claude Code generally available, and shipped the models with measures for its AI Safety Level 3.[9] Anthropic reported that Claude Code passed $1 billion in run-rate revenue in November 2025, about six months after general availability.[18] As of Nov 7, 2026, Claude Code runs in the terminal, in VS Code and JetBrains editors, in the Claude desktop app and in the browser, where it can edit files, run commands, write commits and open pull requests.[32] Most ways of using it require a paid Claude plan or a pay-as-you-go Anthropic Console account, and its default model on every paid plan is Opus 5.5.[32][33]

What Claude Code is Anthropic's Claude channel defines Claude Code as an agent that understands a codebase, edits files and runs commands, lists the places it runs, and explains how it differs from the chat app: it has direct access to the user's files and terminal.

The Claude 5 generation (2026)

On Jun 9, 2026 Anthropic added a tier above Opus with Claude Fable 5, which it described as a model of Mythos class made safe for general use, priced at $10 per million input tokens and $50 per million output tokens.[10] A sibling built on the same base model, Claude Mythos 5, had some safeguards removed and went first to partners in Project Glasswing, a cybersecurity program.[10] On Jun 12 the US government applied export controls to both models that required restricting access by foreign nationals; because Anthropic had no reliable way to check nationality in real time, it suspended both for every user.[11] The directive followed a report by Amazon researchers describing a way around Fable 5's safeguards that led it to find software vulnerabilities.[11] Fable 5 returned on Jul 1, 2026 behind a new classifier, developed with the government, that Anthropic said blocks the reported technique in more than 99% of cases but flags more harmless coding requests.[11] Fable 5.1 and Mythos 5.1 followed in September 2026, Mythos 5.1 only for vetted US organizations in cybersecurity and life sciences.[12] Claude Opus 5.5 arrived on Sep 22, 2026 at $4 and $20 per million tokens, below the $5 and $25 of Opus 5.[13]

Anthropic launches Claude Fable 5 Anthropic's launch film for Fable 5 explains why the earlier Claude Mythos Preview was never broadly released: it found thousands of security flaws, so Anthropic gave it to defenders of critical software first and later released Fable 5 with safeguards for everyone.

Models and pricing

As of Nov 7, 2026, Anthropic's current models are Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5, all with a context window of one million tokens, roughly 555,000 words, and up to 128,000 tokens of output.[1] Anthropic's documentation advises starting with Opus 5.5 for most workloads and turning to Fable 5.1 for demanding reasoning and long agentic work.[1] On the Claude API, input costs $10 per million tokens on Fable 5.1, $4 on Opus 5.5, $2 on Sonnet 5.5 and from $0.10 on Haiku 5.5; output costs five times the input price on each, and batch requests cost half.[1] All four share a reliable knowledge cutoff of June 2026, and Anthropic commits not to retire any of them on its own platforms before September 2027.[1] Older models, including Fable 5, Opus 5 and Haiku 4.5, remain on the API as legacy models.[1] In the apps, the Free plan offers Sonnet and Haiku, Pro adds Opus, and Max and Ultra include Fable for up to half of their weekly limit.[2]

Which model to pick Anthropic's Claude channel advises Fable 5.1 for open-ended work, starting at medium effort, and Opus or Sonnet for routine tasks, since a cheaper model often gets a similar result without paying for capability the task never uses.

Apps and features

Claude's apps run in a web browser at claude.ai, on macOS and Windows, on Linux in beta, and on iOS.[17] In a chat, a user can upload up to 20 files of up to 500 MB each, including PDFs of up to 1,000 pages, although Claude analyzes the charts and pictures in a PDF only when it has 100 pages or fewer.[15] Instead of generated images, Claude builds diagrams, charts and interactive visuals in HTML and SVG, and it can describe pictures a user uploads.[14] Projects keep related conversations together with shared context and custom instructions, and artifacts turn a request into an app, tool or document that can be shared.[31] Anthropic has been merging Cowork, its agent for longer knowledge work, into ordinary chat, first on the Pro, Max and Ultra plans: Claude picks its own tools in a single conversation, and longer tasks keep running in the cloud after the user closes the laptop.[16] Accounts that receive the merged version cannot switch back to separate chat and Cowork modes.[16]

Plans, limits and data

As of Nov 7, 2026, Claude is free to use with limits; Pro costs $20 a month, or $17 a month billed annually, and includes Claude Code; Max costs $100 for five times Pro's usage; and Ultra costs $200 for ten times Pro's weekly usage with no five-hour limit.[2] Anthropic states usage as multiples of Pro rather than as a number of messages.[2] Weekly limits began on Aug 28, 2025, aimed at subscribers who ran Claude Code nonstop in the background and at people who shared or resold accounts; Anthropic estimated that fewer than 5% of subscribers would be affected.[19] On May 6, 2026, after agreeing to use all the computing capacity of SpaceX's Colossus 1 data center, Anthropic doubled Claude Code's five-hour limits on Pro, Max, Team and seat-based Enterprise plans.[30] Since Aug 28, 2025, Free, Pro and Max users have chosen whether their chats and coding sessions may be used to train Anthropic's models; with training on, data is kept for five years instead of 30 days.[20] Team, Enterprise, education, government and API use are excluded from that change.[20]

Safety approach

Anthropic trains Claude on a written constitution that sets out the values and behavior it intends for the model; the current version was published on Jan 22, 2026 and released into the public domain under a CC0 license.[21][22] When goals conflict, it asks Claude to be broadly safe first, then broadly ethical, then compliant with Anthropic's more specific guidelines, then genuinely helpful, weighing these as a whole rather than as a strict ranking.[22] The document also says Anthropic is uncertain whether Claude has some kind of consciousness or moral status, and that the company cares about Claude's wellbeing.[22] Releases are governed by Anthropic's Responsible Scaling Policy, first published in September 2023 and at version 3.4 since Jul 8, 2026, which ties stronger security and deployment safeguards to AI Safety Levels as models become more capable.[34] Claude Opus 4 and Sonnet 4 shipped with measures for the third of those levels in May 2025.[9] The versions of Anthropic's top model with some safeguards removed, Mythos 5 and Mythos 5.1, are offered only to vetted partners.[10][12]

How the scaling policy began In a conversation among Anthropic's co-founders, one recalls discussing the idea with Paul Christiano in late 2022: rather than one cap on scaling, a series of capability thresholds, each requiring tests and stronger safety and security measures, ideally adopted by more than one company.

Reception

On the WebDev Arena leaderboard, where people vote between web apps built by competing models, Claude Opus 5.5 ranked first with a score of 1814 on Nov 6, 2026, ahead of Claude Sonnet 5.5 and OpenAI's GPT-6 Astra.[27] On the Text Arena leaderboard for general chat the same day, Opus 5.5 ranked fourth, behind Google's Gemini 4 Carbon and Gemini 4 Argon and Meta's Muse Wonder.[28] Zapier's comparison, last reviewed in October 2026, judged Anthropic's and OpenAI's flagship models "essentially at parity", recommending Claude to people focused on coding and sophisticated creative work and ChatGPT to those who want an all-in-one toolkit.[29] Commercially, Anthropic's IPO filing, as reported by CNBC, showed revenue of $4.59 billion in 2025 and $16.29 billion in the first half of 2026.[24] In that half year, 80% of revenue was consumption-based and 20% came from subscriptions, and 48% was sold through Amazon Web Services, Google Cloud and Microsoft Azure.[24]

A reviewer's verdict: use both Closing his June 2026 comparison, The AI Advantage's host names Claude the overall winner for people moving into longer, agent-style work, but advises many viewers to use both: ChatGPT for images, Claude for writing and agentic tasks.

Criticism and controversies

The most cited gap is media: Claude makes no images or video, while ChatGPT generates images.[14][29] In a class action by authors, Judge Alsup ruled in June 2025 that training Claude on books was fair use but that downloading pirated copies from Library Genesis and Pirate Library Mirror was not.[25] Anthropic agreed in September 2025 to pay at least $1.5 billion, about $3,000 for each of roughly 500,000 works, and the court granted final approval on Jul 20, 2026.[25][26] In February 2026 Anthropic refused the US Department of War's demand that AI companies accept "any lawful use" of their models, keeping two exceptions: mass domestic surveillance and fully autonomous weapons.[23] The Pentagon then labeled Anthropic a supply chain risk and President Trump ordered federal agencies to stop using its models; a San Francisco federal judge found the actions retaliatory, the D.C. Circuit sided with the Pentagon on the designation, and both rulings were being challenged as of Oct 26, 2026.[24] CNBC also reported that the Federal Trade Commission had demanded testimony from Anthropic executives in an inquiry into possible consumer dangers from AI products.[24]

The round Claude sits out In an August 2026 test of the three $20 plans, Parker Prompts notes that Claude can describe an uploaded image in detail but cannot create one, so the image round is left to ChatGPT and Gemini.

Timeline

  1. Claude and Claude Instant are released.[3]
  2. Claude 2 and the public claude.ai beta open in the US and UK.[4]
  3. Claude Pro launches at $20 a month.[5]
  4. The Claude 3 family introduces Haiku, Sonnet and Opus.[6]
  5. Computer use enters public beta.[7]
  6. Claude 3.7 Sonnet and a preview of Claude Code.[8]
  7. Claude 4 arrives; Claude Code becomes generally available.[9]
  8. Weekly limits and the training opt-out take effect.[19][20]
  9. A new constitution is published.[21]
  10. Claude Fable 5 and Mythos 5 are released; both are suspended Jun 12 to Jul 1.[10][11]
  11. Claude Opus 5.5 is released.[13]

What we think

This section is Heartipedia’s own opinion, not sourced fact. Each view is reasoned from the facts cited above and from trends we see; it may turn out wrong. A prediction’s words are never changed once made: when the facts are known we add the result under it, right or wrong. Every prediction we have made.

Claude 6 launches as a public model and a restricted twin

We expect Anthropic's first Claude 6 model to arrive before the end of 2027 as a pair, a public Fable model and a Mythos model with some safeguards lifted for vetted organizations only, as with both Fable releases before it.

Why: Anthropic released Fable 5 with Mythos 5 on the same base model, then Fable 5.1 with Mythos 5.1, and kept the Mythos versions for vetted partners in cybersecurity and life sciences. A company that held back its strongest model even after an export-control suspension has settled on that split, and new generations have come about a year apart since Claude 4 in May 2025.

Prediction Confidence: Possible Check by: Dec 2027 Made

Images come from someone else's model

We expect the Claude apps to start generating pictures before the end of 2027 by handing the request to another company's image model, not one Anthropic trains itself.

Why: Claude makes no images today and builds charts and diagrams in code instead, while ChatGPT generates images and reviewers keep naming the gap. Every release on the page has added text, code, computer use or agent abilities, never media, and Claude already reaches outside tools through connectors and the Model Context Protocol, the cheapest way to close a gap Anthropic has shown no wish to train for.

Prediction Confidence: Long shot Check by: Dec 2027 Made

The agent comes to the Free plan

We expect Anthropic to give Free users a limited version of the agent mode that came from Cowork, where Claude works through a longer task and keeps going in the cloud, before the end of 2027.

Why: Anthropic has already merged Cowork into ordinary chat on Pro, Max and Ultra, so the agent is now simply how Claude answers on paid plans rather than a separate product. Keeping Free users on a different kind of Claude gets harder to explain each time the paid version changes, and the extra computing capacity Anthropic bought in 2026 is what such a change would need.

Prediction Confidence: Possible Check by: Dec 2027 Made

The next Opus costs $3 or less

We expect the next Opus model after Opus 5.5 to cost $3 or less per million input tokens on the Claude API.

Why: Opus 5.5 came in at $4 per million input tokens, below the $5 of Opus 5, and Anthropic's own launch post presents the lower cost as a headline feature. Sonnet 5.5 already sells at $2, so a cut to $3 would still keep Opus clearly above it, and rivals at the top of the leaderboards compete on price as much as on scores.

Prediction Confidence: Possible Check by: Dec 2027 Made

Claude's limits hurt it more than its models do

We think Claude's biggest weakness is not the quality of its models but how much of them a subscriber is allowed to use, and how hard that is to know in advance. We think a reader who wants one assistant for everything, with images, the most free use and the widest set of features, is better served by ChatGPT today.

Why: A Claude model led WebDev Arena on Nov 6, 2026, and Zapier rates the two flagship families "essentially at parity". Yet Anthropic states its plans only as multiples of Pro, added weekly limits in 2025, and doubled Claude Code's five-hour limits only after buying a whole data center's capacity in May 2026. On general chat Claude is not ahead: Opus 5.5 was fourth on Text Arena, behind two Gemini models and one from Meta.

Opinion Made

Moments

  1. A tour of the Claude app Anthropic's own walkthrough of claude.ai: the sidebar for new and earlier chats, projects that keep conversations with shared context and instructions, and artifacts that turn an idea into an app, tool or document to share.
  2. Computer use fills in a form In Anthropic's October 2024 demonstration, Claude takes screenshots, notices that a company is missing from a spreadsheet, looks it up in a CRM, scrolls for the details and copies them into a vendor form before submitting it, with no help from the person watching.
  3. What Claude Code is Anthropic's Claude channel defines Claude Code as an agent that understands a codebase, edits files and runs commands, lists the places it runs, and explains how it differs from the chat app: it has direct access to the user's files and terminal.
  4. Anthropic launches Claude Fable 5 Anthropic's launch film for Fable 5 explains why the earlier Claude Mythos Preview was never broadly released: it found thousands of security flaws, so Anthropic gave it to defenders of critical software first and later released Fable 5 with safeguards for everyone.
  5. Which model to pick Anthropic's Claude channel advises Fable 5.1 for open-ended work, starting at medium effort, and Opus or Sonnet for routine tasks, since a cheaper model often gets a similar result without paying for capability the task never uses.
  6. How the scaling policy began In a conversation among Anthropic's co-founders, one recalls discussing the idea with Paul Christiano in late 2022: rather than one cap on scaling, a series of capability thresholds, each requiring tests and stronger safety and security measures, ideally adopted by more than one company.
  7. A reviewer's verdict: use both Closing his June 2026 comparison, The AI Advantage's host names Claude the overall winner for people moving into longer, agent-style work, but advises many viewers to use both: ChatGPT for images, Claude for writing and agentic tasks.
  8. The round Claude sits out In an August 2026 test of the three $20 plans, Parker Prompts notes that Claude can describe an uploaded image in detail but cannot create one, so the image round is left to ChatGPT and Gemini.

References

  1. Models overview, Anthropic, 2026-11-07primary
  2. Plans & pricing, Anthropic, 2026-11-07primary
  3. Introducing Claude, Anthropic, 2023-03-14primary
  4. Claude 2, Anthropic, 2023-07-11primary
  5. Introducing Claude Pro, Anthropic, 2023-09-07primary
  6. Introducing the next generation of Claude, Anthropic, 2024-03-04primary
  7. Introducing computer use, a new Claude 3.5 Sonnet, and Claude 3.5 Haiku, Anthropic, 2024-10-22primary
  8. Claude 3.7 Sonnet and Claude Code, Anthropic, 2025-02-24primary
  9. Introducing Claude 4, Anthropic, 2025-05-22primary
  10. Claude Fable 5 and Claude Mythos 5, Anthropic, 2026-06-09primary
  11. Redeploying Claude Fable 5, Anthropic, 2026-06-30primary
  12. Introducing Claude Mythos and Fable 5.1, Anthropic, 2026-09primary
  13. Introducing Claude Opus 5.5, Anthropic, 2026-09-22primary
  14. Can Claude produce images?, Anthropic, 2026-11-07primary
  15. Upload files to Claude, Anthropic, 2026-11-07primary
  16. Claude Cowork and chat are one Claude, Anthropic, 2026-11-07primary
  17. Download Claude, Anthropic, 2026-11-07primary
  18. Anthropic acquires Bun as Claude Code reaches $1B milestone, Anthropic, 2025-12-03primary
  19. Anthropic unveils new rate limits to curb Claude Code power users, TechCrunch, 2025-07-28press
  20. Updates to Consumer Terms and Privacy Policy, Anthropic, 2025-08-28primary
  21. Claude's new constitution, Anthropic, 2026-01-22primary
  22. Claude's Constitution, Anthropic, 2026-11-07primary
  23. Statement from Dario Amodei on our discussions with the Department of War, Anthropic, 2026-02-26primary
  24. Anthropic IPO live updates, CNBC, 2026-10-26press
  25. AI company Anthropic agrees to pay $1.5B to settle lawsuit with authors, NBC News, 2025-09-05press
  26. Anthropic Copyright Settlement, Settlement Administrator, 2026-11-07primary
  27. WebDev Arena leaderboard, Arena, 2026-11-06primary
  28. Text Arena leaderboard, Arena, 2026-11-06primary
  29. Claude vs. ChatGPT: What's the difference?, Zapier, 2026-10press
  30. Higher usage limits and a SpaceX compute deal, Anthropic, 2026-05-06primary
  31. Getting started with Claude.ai, Anthropic, 2025-12-02primary
  32. Claude Code overview, Anthropic, 2026-11-07primary
  33. Model configuration, Anthropic, 2026-11-07primary
  34. Responsible Scaling Policy, Anthropic, 2026-08-14primary

Reference links open in a new tab, so this page stays where you left it.

About this page

Drafted by
claude-opus-5-5 on 2026-11-07T12:00:00Z
Checked by
no AI check has run on this page
Reviewed by
not yet reviewed by a named human
Status
Published and revised continuously; this version is the one described here.
Published
7 November 2026
Last updated
7 November 2026, see what changed
Sources
No Wikipedia source; the sources are listed above.
License
Text CC BY-SA 4.0; embedded videos and thumbnails remain their uploaders’ property, shown under YouTube’s Terms of Service; quotations excluded. Details

Suggest a change Revision history

Keyboard shortcuts

/
Jump to the search box
F
Full screen on or off
Ctrl + ,
Settings
? or Ctrl + /
Open this list
Esc
Close this list, or the Settings or account menu

Video, on a page that has one

Space or K
Play or pause
← →
Back or forward 5 seconds
J L
Back or forward 10 seconds
0 to 9
Go to 0% to 90% of the video
, .
Previous or next frame, while paused
M
Mute or unmute
C
Captions on or off
F
The video full screen on or off
A D
Previous or next moment