What is Claude and how it can help you be more productive

Anthropic assistant guide · Updated August 2026

Claude isn’t just Anthropic’s cautious answer to ChatGPT anymore. What launched in March 2023 as a safety-focused chatbot from a lab founded by ex-OpenAI researchers has grown into a full agentic platform: a conversational assistant, a terminal-based coding agent, a desktop coworker that reaches into your files, and a shared teammate that lives inside Slack. In July 2026 Anthropic shipped Claude Opus 5 — its fourth model release in under two months.

In this guide we explain what Claude is in 2026, how the Fable 5, Mythos 5, Opus 5, Sonnet 5, and Haiku 4.5 model family works, what Claude Code, Cowork, and Claude Tag can do, what the plans cost, and how to use Claude well.

Mar 2023First public release, Claude 1 via Anthropic API
Fable 5 · Opus 5Current frontier tier and flagship
Code · Cowork · TagAgentic surfaces across terminal, desktop, and Slack
01 · What it is

What is Claude?

Claude is Anthropic’s family of AI models, along with the assistant products built around them. Anthropic was founded in January 2021 by siblings Dario Amodei (CEO) and Daniela Amodei (President), together with five other researchers who left OpenAI over how much weight safety should carry in frontier AI development. The company is structured as a Public Benefit Corporation, and its signature training method, Constitutional AI, steers models to follow a written set of principles rather than relying purely on human raters to shape behavior after the fact.

Claude reached the public on March 14, 2023, as Claude 1 and Claude Instant, offered through Anthropic’s API before ChatGPT had even marked its first birthday. What has changed since then is almost everything around that original model: extended thinking, a 1-million-token context window, computer use, native file creation, memory, dozens of connectors, and three separate agentic products that let Claude act instead of only advise.

Today, depending on the plan and the surface, Claude can:

  • Answer questions & explain concepts
  • Search the live web & cite sources
  • Reason step by step with extended thinking
  • Write and edit documents in Artifacts
  • Create Excel, PowerPoint & PDF files
  • Analyze data and run code
  • See and control a computer
  • Write, test & ship code as an agent
  • Remember context via Memory
  • Work across Slack, Drive, Gmail & more
Claude – the models

Claude is Anthropic’s family of large language models. Sizes are named rather than numbered by parameter count: Haiku (fastest and cheapest), Sonnet (balanced), and Opus (most capable) — joined in June 2026 by a new tier above Opus, called Mythos.

Claude – the products

Claude.ai (web, iOS, Android, desktop) is the most visible surface, but the same models sit underneath Claude Code, Claude Cowork, Claude Tag, Claude in Chrome, the Excel/PowerPoint add-ins, and the Developer Platform via Anthropic API, Amazon Bedrock, and Vertex AI.

02 · The model family

Claude 5: Fable, Mythos, Opus, Sonnet, and Haiku

Claude’s current generation carries the “5” name for its two newest tiers — Sonnet 5 (June 30, 2026) and Opus 5 (July 24, 2026) — while a Mythos-class tier above Opus, made up of Claude Fable 5 and Claude Mythos 5, launched a few weeks earlier on June 9, 2026. As with rivals, Claude.ai mostly assigns a model for you by plan, with a picker when you want to choose by hand.

Frontier tier · Mythos-class · June 9, 2026

Fable 5 & Mythos 5

Anthropic’s most capable released models, sharing one architecture. Fable 5 ships with extra safeguards around biology, cybersecurity, and LLM R&D, which is why it’s the public version; Mythos 5 drops that layer and is offered only to vetted partners via Project Glasswing for defensive cybersecurity work.

Maximum capabilityFable public · Mythos invite-only
Near-frontier · Opus tier · July 24, 2026

Opus 5

The newest Opus release, close to Fable 5’s intelligence at roughly half the price. Default on Claude Max and the strongest model on Pro, tuned for agentic coding, computer use, and long-horizon reasoning, with low/medium/high effort control per request.

Best cost/capability at the top
Everyday default · Sonnet tier · June 30, 2026

Sonnet 5

Anthropic’s most agentic Sonnet yet, with performance close to Opus 4.8 at a fraction of the cost. The default model on Free and Pro and the practical choice for most production agent workloads.

Best value for everyday work
Fastest & cheapest · Oct 15, 2025

Haiku 4.5

The quickest, lowest-cost model, and the first Haiku to ship with extended thinking, computer use, and awareness of its own remaining context. Near Sonnet 4 intelligence at about a third of the cost.

Lowest cost per task

Behind this sit the still-supported Opus 4.8, 4.7, 4.6, and Sonnet 4.6/4.5 — each only months old but not yet retired, since API customers often stay a generation behind to avoid re-testing production workflows.

03 · How it works

How does Claude work?

Claude models are trained with Constitutional AI: instead of shaping behavior purely through human feedback, Anthropic has models critique and revise their own draft responses against a written set of principles, then reinforces that behavior during training. Alongside sits the Responsible Scaling Policy and AI Safety Levels (ASL) — Opus 4 was the first Claude deployed under the stricter ASL-3 standard, and Fable 5’s extra biology/cybersecurity safeguards are the clearest recent example.

Modern Claude combines the model with tools and adaptive reasoning:

  • 1

    Understand the request

    Claude reads your prompt, conversation so far, uploads, connectors, and Memory.

  • 2

    Reason through the problem

    Extended thinking (since 3.7 Sonnet Feb 2025) lets Claude work step by step before answering. An effort setting (low/medium/high) now controls reasoning compute per request.

  • 3

    Decide which tools to use

    Web search, code execution, file generation, Computer Use screenshots, or MCP connectors (open standard Anthropic released Nov 2024).

  • 4

    Generate the response or take the action

    The answer, spreadsheet, pull request, or finished deliverable — sometimes after working independently for minutes or hours.

  • 5

    Refine with follow-ups

    Artifacts and Projects keep documents and code open and editable across the thread instead of restarting from scratch.

04 · What you can do

What can you do with Claude?

Research and information

Web search grounds answers in current results with citations instead of only training data.

Writing and editing

Draft, rewrite, summarize, and translate. Artifacts keep a document or webpage open in its own pane so you can iterate, and Projects hold files and instructions for recurring work.

Coding and DevOps

Claude Code turns Claude into a terminal engineering collaborator that reads a real codebase, edits files, writes and runs tests, and manages Git — estimated $2.5B annualized revenue within a year of GA.

Files and office work

Create and edit real Excel, PowerPoint, Word, and PDF files directly, and via add-ins work inside Microsoft apps, passing context between spreadsheet and deck without copy-paste.

Data analysis

Upload spreadsheets or CSVs, let Claude clean data, run calculations in a sandbox, and explain what changed instead of just handing back numbers.

Computer use and browsing

Claude can see a screen, move a cursor, click, and type — behind Cowork’s desktop work and Claude in Chrome’s scoped browser automation (rolled out cautiously due to prompt-injection risk).

Memory and personalization

Recall relevant context from earlier chats so you don’t restate it, with incognito chats when you’d rather not be remembered.

Enterprise integrations

Connectors via MCP reach into Drive, Gmail, Slack, GitHub and dozens of tools, plus verticals like Financial Services and Life Sciences for regulated industries.

05 · From assistant to teammate

The agentic shift: Claude Code, Cowork, and Claude Tag

Anthropic’s biggest change since 2023 isn’t a smarter chatbot — it’s that Claude increasingly does the work instead of only describing how to do it. Three products carry that shift, each for a different way of working.

01

Claude Code

Anthropic’s original agentic product, research preview Feb 24, 2025 and GA May 22, 2025. It runs in a terminal, searches and edits a real codebase, writes and runs tests, and commits changes — keeping a developer in the loop at each step. By early 2026 estimated over half of enterprise coding-model workloads and a meaningful share of public GitHub commits.

Included with Pro, Max, Team Premium & Enterprise

02

Claude Cowork

Research preview Jan 12, 2026 for Max on macOS, then expanded to Pro/Team/Enterprise within weeks and to Windows Feb 10. Cowork brings Code’s agent loop to everyday knowledge work: point it at a folder, describe an outcome — expense report from receipt photos, say — and it plans, works, and returns a finished result, coordinating multiple sub-agents on complex jobs.

Included with every paid plan, not Free

03

Claude Tag

Beta June 23, 2026, a persistent shared Claude inside Slack channels. Where Code and Cowork are single-player, Tag is deliberately “multiplayer”: one Claude per channel any member can tag into, pick up mid-flight, or redirect, with optional ambient mode that flags things unprompted.

Beta · Team & Enterprise plans

Claude in Chrome and the Excel/PowerPoint add-ins extend the same idea into browser and Microsoft apps, while Claude Design (research preview April 2026) applies it to slides and prototypes.

06 · Key moments

Claude timeline: 2021–2026

Jan 2021

Dario and Daniela Amodei found Anthropic as a Public Benefit Corporation with five other ex-OpenAI colleagues, around Constitutional AI.

Mar 2023

Claude 1 and Claude Instant launch via Anthropic API — first public release. 100K-token window arrives that May, ahead of any frontier lab.

Jul–Nov 2023

Claude 2 ships; 2.1 doubles context to 200K in November.

Mar 2024

Claude 3 introduces Haiku/Sonnet/Opus structure with vision in all three tiers, carried through every generation since.

Jun 2024

Claude 3.5 Sonnet launches alongside Artifacts, the live editing pane.

Oct 2024

Updated 3.5 Sonnet and 3.5 Haiku ship with Computer Use in public beta.

Nov 2024

Anthropic open-sources Model Context Protocol (MCP), the standard for connecting assistants to external tools.

Feb 2025

Claude 3.7 Sonnet introduces hybrid extended-thinking reasoning; Claude Code launches as research preview the same day.

May 2025

Claude Opus 4 and Sonnet 4 reach GA alongside Claude Code’s own GA — Opus 4 first Claude under ASL-3.

Sep–Oct 2025

Sonnet 4.5 adds context-awareness; Haiku 4.5 follows in October as first Haiku with extended thinking and computer use.

Nov 2025

Opus 4.5 ships alongside Chrome and Excel integrations.

Jan–Feb 2026

Cowork rolls out from Max to every paid plan; Opus 4.6 and Sonnet 4.6 follow, with 1M context window in beta.

May 2026

Opus 4.8 launches.

Jun 2026

Fable 5 and Mythos 5 launch as first Mythos-class models (June 9); export-control suspension June 12–30; Tag launches in Slack June 23; Sonnet 5 launches June 30.

Jul 2026

Access to Fable 5 restored July 1; Opus 5 launches July 24 — fourth model release in under two months.

07 · Staying productive

How to use Claude well

Be specific

State audience, goal, format, and constraints up front. “Turn this into 5 bullet points for a board that’s never seen the data” beats “summarize this.”

Use Artifacts and Projects

Keep a document or code open and editable instead of restating it every turn; Projects hold files and instructions so every new chat starts with the right context.

Pick the right surface

Quick answer: normal chat. Current facts: web search. Code: Code. Office files and folder work: Cowork. Shared team workflows: Tag. Match surface to job.

The prompt pattern that works

Role:You are a senior analyst writing for a board that hasn’t seen this data before.Task:Summarize our Q2 results and flag the three biggest risks.Format:One page, bold key numbers, risks as bullets with one line of evidence each.

Role, task, format — the same structure every assistant in this series responds well to. For long-running work, put background material in a Project once instead of re-pasting it into every chat.

08 · The window

How large is the context window?

Claude Fable 5, Mythos 5, Opus 5, and Sonnet 5 all carry a standard 1 million-token context window — roughly 750,000 words — with up to 128,000 tokens of output. Haiku 4.5 stays at 200,000 tokens, about 150,000 words, with a 64,000-token output ceiling.

Fable 5 / Mythos 5 / Opus 5 / Sonnet 5 – 1,000,000 tokens≈ 750,000 words
Haiku 4.5 – 200,000 tokens≈ 150,000 words
Claude 3.5 Sonnet (2024) – 200,000 tokens≈ 150,000 words
Claude 1 (2023 launch) – 9,000 tokens≈ 6,750 words

Practical rule of thumb: a token is roughly 0.75 English words, and about 1.5 tokens per word once you factor in punctuation and spacing. Claude was the first frontier-lab model to reach a 100,000-token window in May 2023.

09 · Plans and pricing

What does Claude cost?

PlanPriceModel accessLimits
Free$0Sonnet 5Capped daily messages; no Code or Cowork
Pro$20/mo ($17/mo annual)Sonnet 5 default, Opus 5 available5x Free’s usage; first plan to include Cowork
Max 5x$100/moOpus 5 default5x Pro’s usage per session, for sustained Code/Cowork
Max 20x$200/moOpus 5 default20x Pro’s usage, for parallel sessions and heavy agents
Team Standard$20/seat/mo annual ($25 monthly)Sonnet 5, Opus 52–150 seats; shared billing and admin
Team Premium$100/seat/mo annual ($125 monthly)Adds Code, Cowork, Fable 5 in limitsSame 2–150 seats
Enterprise$20/seat/mo + usage at API ratesFull lineupSSO, SCIM, HIPAA-eligible; 20+ seats self-serve, 50+ sales-assisted
APIusage-basedFable 5 $10/$50 · Opus 5 $5/$25 · Sonnet 5 $2/$10 · Haiku 4.5 $1/$5 per 1MPrompt caching -90%, Message Batches -50%

Sonnet 5 $2/$10 intro made permanent Aug 10, 2026. Prices and limits change often; the chart above reflects August 2026.

10 · FAQ

Frequently asked questions

Is Claude free?

Yes. Free gives you Sonnet 5 with capped daily use and no Code or Cowork. Pro from $17–20 a month removes most ceiling and is the first plan to include Cowork.

Is Claude better than ChatGPT or Gemini?

Anthropic leads on coding and enterprise agents — estimates put Claude around 40% of enterprise LLM spend and over half of enterprise coding-model market — and Code, Cowork, and Tag give it arguably the deepest agentic lineup. ChatGPT and Gemini still reach larger consumer audiences and ship native image/video generation that Claude doesn’t match. For professional and coding work Claude is a strong default.

Does Claude have an app?

Yes — iOS, Android, macOS, Windows, web at claude.ai, Chrome extension, Excel/PowerPoint add-ins, and Tag inside Slack.

What happened with Fable 5 and Mythos 5?

Anthropic released both June 9, 2026 as its first Mythos-class tier above Opus. Three days later the US Commerce Department issued an export-control directive and Anthropic suspended both globally because it had no reliable nationality verification. Controls were lifted June 30 and Fable 5 was fully restored July 1; Mythos 5 remains limited to vetted partners via Project Glasswing.

What are Claude’s limitations?

It can still hallucinate on obscure or fast-moving facts, so verify anything critical. No native real-time voice or built-in image/video generation like ChatGPT and Gemini. Cowork and Claude in Chrome are still previews with real prompt-injection risk when an agent has file or browser access — review any agent action that touches real systems, money, or production code.

Is my data used for training?

Since Aug 2025, Anthropic can use Free, Pro, and Max conversations to improve models unless you turn this off in settings; incognito chats aren’t retained for training. Team, Enterprise, and standard API usage are excluded by default.

Heads-up:Claude is a fast-moving product, and Anthropic has shipped four model generations and three agentic surfaces in the past 18 months alone. Plan names, model names, and prices change frequently — the details above reflect August 2026. Check Anthropic’s official pricing and docs for current limits.

Related reading