Anthropic

Claude

Provider

Anthropic

Price

$0-200/mo

Platforms

Web, iOS, Android

Weekly users

7,500,000

Launched

Mar 2023

Updated August 2026

Claude is Anthropic’s flagship AI assistant and the product that established the company as a genuine third force in frontier AI alongside OpenAI and Google. First released in March 2023, Claude has grown from a safety-focused chatbot into the leading AI for software development — its current flagship, Claude Opus 5 (released 24 July 2026), tops the independent Artificial Analysis Intelligence Index — and, in 2026, into a genuine work platform rather than just a chat box. Anthropic is now the most valuable AI startup in the world at a $965 billion valuation, and much of its growth runs through Claude itself.

The big change in 2026 is what Claude does. Beyond chat, the app now includes Cowork, a desktop agent that completes multi-step work on your own files and apps, and Claude Design, a tool for generating editable visuals, decks and prototypes — plus upgraded voice on both desktop and mobile. This guide focuses on the app most people actually use: the free, Pro and Max experience across web, desktop and mobile.

Quick stats

Pricing breakdown

Claude offers five tiers, each targeting different use cases and budgets. All prices are in USD.

Free — $0

The free tier provides genuine utility for light users. You get:

• Claude Sonnet 5 — Default model for free users, Anthropic’s most agentic Sonnet
• Claude Haiku 4.5 — Fast, efficient model for quick tasks
• ~30-40 messages per 5-hour window — Resets every 5 hours
• Web search — Search the internet for current information
• File uploads — Up to 20 files per chat, 30MB per file
• Image analysis — Vision capabilities for uploaded images
• Artifacts — Code previews and document generation
• Voice mode — ~20-30 voice messages per session
• 5 Projects — Workspaces with persistent context

Limitations: No Opus-tier access (Opus 5 or Opus 4.8), no Memory, no Cowork or Claude Design, no Claude Code, no Google Workspace connectors, stricter rate limits than paid tiers.

Worth it for: Casual users, students, anyone wanting to try Claude before subscribing, light research tasks.

Pro — $20/month

The core subscription for regular users. Everything in Free, plus:

• All Claude models including Opus 5 — Access to Anthropic’s flagship Opus model
• 5x more usage than Free — ~45 messages per 5-hour window
• Cowork and Claude Design — The desktop agent and the visual-design tool, in research preview (Claude Design is metered separately)
• Unlimited Projects — Create as many workspaces as needed
• Extended thinking — Visible reasoning with customisable thinking budgets
• Memory — Claude remembers preferences and context across chats
• Claude Code — Agentic coding in the terminal and IDE
• Research feature — Deep research capabilities
• Google Workspace connectors — Gmail, Drive, Calendar integration via MCP
• Priority during high traffic

Worth it for: Most regular users — developers, writers, researchers and analysts who want frontier models plus the new Cowork and Design tools without Max-level spend.

Subscribe to Pro →

Max 5x — $100/month

For power users who hit Pro limits regularly. Everything in Pro, plus:

• 5x Pro usage — ~225 messages per 5-hour window
• Higher Cowork and Claude Design allowances — More autonomous-agent and design capacity
• Claude in Chrome — full Cowork sessions in the browser (also on Team, and rolling out to Pro since 12 Aug 2026)
• Early access to new features
• Higher priority than Pro

Worth it for: Heavy daily users and professionals who lean on Cowork and Claude Code through the workday and can’t afford to hit rate limits.

Introduced April 2025 →

Max 20x — $200/month

For the most demanding users. Everything in Max 5x, plus:

• 20x Pro usage — ~900 messages per 5-hour window
• Near-unlimited Opus 5 access — Practical limits rarely reached
• Largest Cowork and Claude Design allowances
• Highest priority access — Never wait during peak times

Worth it for: Professional developers, researchers and consultants who use Claude as their primary work tool for several hours daily. Comparable to ChatGPT Pro at the same price point.

Usage limits changed in 2026. Anthropic doubled the 5-hour limits for Pro, Max, Team and Enterprise on 6 May 2026, and raised weekly limits by 50% through 13 July 2026. Max plans also carry weekly caps (one across all models, one for Sonnet). From 15 June 2026, heavy programmatic usage (scripts, the Agent SDK, third-party agent apps) draws on a separate monthly credit pool billed at API rates — the in-app chat, Cowork and Design experience is unaffected.

Team — $30/user/month

Designed for organisations (minimum 5 seats):

• All Pro features
• SSO and admin controls — Centralised user management
• Shared workspaces — Collaborate on Projects across team members
• Microsoft 365 and Slack integrations
• Data never used for training — Explicit exclusion policy
• Premium Team seats available — $150/user/month includes Claude Code

Learn more about Team →

Enterprise — Custom pricing

For large organisations requiring maximum security and customisation:

• 500K+ token context windows — Handle massive documents
• SCIM provisioning — Automated user management
• Audit logs — 180 days of activity tracking
• Role-based access controls
• Custom data retention policies
• VPC isolation options — Via AWS Bedrock or Google Vertex AI
• Dedicated support and onboarding

Notable enterprise customers include Salesforce, Netflix, Snowflake ($200M partnership), Novo Nordisk, Cognizant (350,000 employees), and Palo Alto Networks.

Contact sales →

Models powering Claude

Claude moved through several Opus releases in the first half of 2026. In the app you pick a model from a dropdown; here’s what’s current. For full benchmarks and pricing, each model has its own page.

Claude Opus 5 (July 2026)

The current flagship, released 24 July 2026 and the default Opus model in the app:

• 96.0% on SWE-bench Verified (97.0% on vals.ai’s harness) and #1 on the independent Artificial Analysis Intelligence Index (61 — the highest score it has recorded)
• 1M-token context window with extended thinking on by default
• $5 / $25 per million tokens (API) — half the price of the Mythos-class Fable 5
• Caveat: hallucinates factual claims slightly more than Opus 4.8, so verify uncertain answers

Available to: Pro, Max, Team and Enterprise; the default model on Max.

Claude Opus 4.8 (May 2026)

The previous flagship, released 28 May 2026 as a “modest but tangible” upgrade to Opus 4.7 and still available on all paid tiers:

• 88.6% on SWE-bench Verified — and a category-best 69.2% on the memorisation-resistant SWE-bench Pro
• 1M-token context window — process entire codebases or large document sets
• Honesty gains — around four times less likely than Opus 4.7 to let flaws in its own code pass unremarked
• Pricing (API): $5 / $25 per million tokens

Available to: Pro, Max, Team and Enterprise users.

Claude Sonnet 5 (June 2026)

The mid-tier workhorse and the default model on the Free and Pro tiers, released 30 June 2026:

• 85.2% on SWE-bench Verified — Anthropic’s most agentic Sonnet yet
• 63.2% on SWE-bench Pro — within ~6 points of Opus 4.8, ahead of Sonnet 4.6
• 1M-token context, 128K output
• Pricing (API): $2 / $10 per million tokens

Available to: All tiers including free (the default on Free and Pro). Note: a new tokenizer counts more tokens per request, so model your real per-task cost. Sonnet 4.6 and 4.5 remain available via the API.

Claude Haiku 4.5 (October 2025)

The speed model delivering near-frontier quality at a fraction of the cost:

• 73.3% on SWE-bench Verified
• First Haiku with extended thinking and computer use
• 2x+ faster than Sonnet at roughly a third of the cost
• Pricing (API): $1 / $5 per million tokens

Available to: All tiers including free.

Mythos-class — Fable 5 and Mythos 5

In June 2026 Anthropic introduced a new tier above Opus. Claude Fable 5 topped every coding benchmark Anthropic tested. It and its restricted twin Mythos 5 were suspended worldwide on 12 June 2026 under a US export-control directive, but the US Commerce Department lifted the controls on 30 June and Fable 5 returned to general availability from 1 July 2026. It is not part of the core Claude subscription, though: it is reachable in the app on paid tiers through 12 July 2026, but as an add-on paid through usage credits rather than bundled into the plan. Meanwhile Mythos 5 stays restricted to trusted-access partners (Anthropic). As of 24 July 2026, Opus 5 is the default flagship in the app (Opus 4.8 remains available), with Fable 5 the frontier add-on above it.

Earlier Opus releases

• Claude Opus 4.7 (April 2026) — superseded by 4.8; still available via API
• Claude Opus 4.6 (February 2026) — first Opus with a 1M-token context
• Claude Opus 4.5 (November 2025) — the previous flagship

Features deep dive

Cowork — the desktop agent

Cowork is the biggest change to the Claude app in 2026. Launched as a research preview in January 2026 and made generally available on 9 April 2026, it brings Claude Code’s autonomy to non-technical work — through a normal interface, not a terminal.

• Works on your own files and apps — Point Cowork at folders you choose and it can read, edit and create files to deliver finished work, not just instructions
• Goal-in, deliverable-out — Describe an outcome (clean up this folder, turn these notes into a brief, extract terms from these contracts) and Claude does the multi-step work
• Scheduled, recurring tasks — Set a cadence once: triage email each morning, pull weekly metrics, run a Slack digest
• Connectors and plugins — Google Drive, Gmail, Docusign, FactSet and domain plugins for finance, engineering and HR

Now on mobile and web, with cloud execution. On 7 July 2026 Anthropic took Cowork beyond the desktop: it now runs in the sidebar of the Claude app on iOS, Android and the web (no separate app), in beta rolling out from Max subscribers first. The bigger change is under the hood — Cowork’s new surfaces run as remote sessions on Anthropic’s servers rather than only locally on your machine, so a task keeps running in the background even with your laptop closed or offline. You can start a job at your desk, get status updates and permission prompts on your phone, and pick up the finished output later (TechCrunch).

Where it runs: the Claude desktop app (macOS/Windows), plus iOS, Android and web (beta). Available to: all paid plans (Pro, Max, Team, Enterprise) — not the free tier; the mobile/web beta began with Max and is expanding to other plans.

New — teach Claude a skill by recording your screen. On 21 July 2026, Anthropic added Record a skill to Cowork’s + menu in the desktop app. You record your screen while doing a task and talk through the steps as you go; Claude turns the demonstration — screen, clicks, keystrokes and voice narration — into a saved, reusable skill it can run again on its own, with no prompt-writing or coding (the Decoder). It is the “show, don’t tell” way to automate a repeatable workflow, and it is available on Pro, Max and Team plans.

Claude Design — visuals and prototypes

Claude Design launched on 17 April 2026 as a Design tab inside Claude. Describe what you want — a landing page, an app screen, a pitch deck, a marketing one-pager — and it produces a first version in seconds.

• Live, editable output — Results are clickable HTML, not static images; refine them with voice, inline comments or sliders
• Uses your design system — Reads your codebase and Figma files to extract your styles, then applies them to new work (VentureBeat)
• Exports everywhere — Canva, PDF, PPTX, HTML, and a hand-off to Claude Code for developers

Available to: Pro, Max, Team and Enterprise in research preview. It is priced and metered separately from the rest of Claude, with its own weekly allowance and purchasable credits.

Web search

Web search launched for paid US users in March 2025 and expanded globally in May 2025. Powered by Brave Search, it provides:

• Automatic triggering — Claude determines when current information is needed
• Clickable citations — Source links for verification
• Real-time information — News, stock prices, current events

Unlike ChatGPT’s always-on approach, Claude’s web search is selective—it searches only when beneficial rather than for every query. Available on all tiers including free.

Research feature

The Research feature operates as Claude’s deep research capability, autonomously exploring multiple sources to compile comprehensive reports. Similar to ChatGPT’s Deep Research but with Claude’s characteristic thoroughness and citation style.

Extended thinking

Extended thinking launched February 2025 with Claude 3.7 Sonnet, allowing users to see Claude’s reasoning process:

• Visible thought process — Watch Claude work through complex problems
• Customisable thinking budgets — Up to 64,000+ tokens for deep reasoning
• Improved accuracy — Particularly for math, coding, and analysis

Available to Pro, Max, Team, and Enterprise users.

Claude Code

Claude Code is Anthropic’s agentic coding tool, and it has become a breakout product — passing a $2.5 billion annualised run rate by early 2026. For app users it now has a desktop app of its own (redesigned in 2026 for managing parallel agents), so you don’t have to live in a terminal to use it: it reads and edits whole projects, runs commands, and works through coding tasks on its own. It is included with Pro and Max — the same subscription that powers the chat app — and shares your plan’s usage.

This guide stays focused on the everyday Claude app; for the developer-tool depth (terminal, IDE extensions, the Agent SDK), Claude Code is its own product.

Note for developers: Claude Code output is covered by Anthropic’s new watermarking (August 2026) like every other Claude surface — though statistical text watermarks are weakest on code, where formatters and linters rewrite the low-entropy token choices that carry the signal (see our watermarking guide).

Available to: Pro, Max, Team, Enterprise users.

Voice — now on desktop too

Voice started as a mobile feature in 2025 and expanded significantly in 2026, on both desktop and mobile.

Desktop (Quick Entry). The Mac app’s Quick Entry lets you reach Claude from anywhere on your computer without switching windows: press Caps Lock to dictate with real-time transcription, capture a screenshot, or share an application window straight into a chat. It turns Claude into an always-available assistant rather than a tab you open.

Mobile (conversational voice). The phone app’s voice mode was revamped in 2026 with multilingual support (Spanish, German, Portuguese, Chinese, Japanese, Russian and Ukrainian among them) and a push-to-talk option for noisy settings. You can talk naturally, switch between voice and text, and keep web search and file analysis available mid-conversation.

How it works: voice is still built on text-to-speech rather than fully native audio, so it is not as fluid as ChatGPT’s real-time voice — but desktop dictation and multilingual mobile voice make it far more useful day to day than it was.

Available to: all tiers; free users get a limited number of voice messages per session.

Artifacts

Artifacts launched June 2024, creating a dedicated side panel for substantial content:

What Artifacts can display:

• Code with syntax highlighting
• Documents (markdown, HTML)
• Interactive React components
• Mermaid diagrams
• SVG graphics

Key capabilities:

• Version history — Track changes across iterations
• Public publishing — Share artifacts with shareable links
• Remix capability — Others can modify shared artifacts
• Export options — Download as files

Artifacts represent a genuine differentiator—ChatGPT’s Canvas is similar but less mature.

Projects

Projects launched June 2024, providing self-contained workspaces:

• Persistent knowledge bases — Upload files that remain available across all chats
• Custom instructions — Project-specific system prompts
• 200K token context windows — Large document processing
• Collaboration — Team/Enterprise users can share Projects with permission levels

Limits:

• Free: 5 Projects
• Pro/Max: Unlimited Projects
• Team/Enterprise: Unlimited with collaboration features

Memory

Memory rolled out August-October 2025, allowing Claude to remember information across conversations:

• Automatic learning — Claude identifies important details to remember
• Project-scoped — Memory stays within specific Projects for separation
• Import capability — Bring memory from ChatGPT or Gemini
• User control — View, edit, and delete memories

Important: Memory is only available on Pro, Max, Team, and Enterprise plans—not free tier.

MCP and connectors

Model Context Protocol (MCP) launched November 2024 as Anthropic’s open standard for connecting Claude to external tools:

Official integrations:

• Google Workspace (Gmail, Drive, Calendar)
• GitHub
• Slack
• Notion
• Asana
• Figma
• Jira
• And many more via community servers

MCP has gained industry adoption—OpenAI announced support, Microsoft integrated it into developer tools. This positions Claude as a hub for enterprise workflows.

Claude in Chrome — now a Cowork session

Claude in Chrome puts Claude in a browser side panel where it can see the page you’re on and act in it — clicking links, typing text, navigating between pages and filling out forms, using your existing logins. On 12 August 2026 Anthropic made the side panel a full Cowork session, which changes what it’s for:

• Conversations are saved to your history — browser sessions are no longer throwaway
• Skills and connectors work in the browser — the same ones you use in Cowork
• Cross-device hand-off — a task started in a tab can be finished in the Claude desktop, web or mobile apps (Anthropic’s example: collect invoices from a vendor portal in Chrome, then build the spreadsheet on desktop)
• Safety checks — before consequential actions (submitting a form, sending a message, downloading a file) a separate check reviews the action against what you originally asked; Claude still asks before irreversible or costly actions like purchases, and Anthropic acknowledges prompt injection remains a risk it mitigates with each model release rather than eliminates

Availability: Max and Team plans now, rolling out to Pro over the coming weeks — no longer Max-only. Enterprise has it off by default, with admin enablement and approved-domain restrictions. Not yet on other Chromium browsers or mobile (9to5Mac).

Computer use

Computer use allows Claude to control desktop applications:

• Screen reading — Understand what’s on screen
• Mouse and keyboard control — Interact with applications
• 61.4% on OSWorld — Sonnet 4.5 leads all models

Currently available via API and Claude Code, with broader consumer rollout expected.

File uploads and handling

Claude supports extensive file upload capabilities:

Supported formats: PDF, DOCX, XLSX, CSV, TXT, HTML, PNG, JPEG, GIF, WebP, code files

Limits:

• Up to 30MB per file
• Up to 20 files per chat
• Text content processed within 200K token context

Strengths

Coding excellence — Claude leads independent coding benchmarks. Opus 4.8 scores 88.6% on SWE-bench Verified and a category-best 69.2% on the harder, memorisation-resistant SWE-bench Pro, ahead of GPT-5.5 and Gemini 3.1 Pro. It remains the default choice in many developer tools.

Real work, not just chat — With Cowork and Claude Design, the app now completes multi-step tasks on your files and generates editable visuals — a genuine step beyond a question-and-answer chatbot.

Writing quality — Users consistently praise Claude’s natural, less robotic writing. It avoids marketing jargon and template responses better than most rivals, and is particularly strong for long-form content, analysis and nuanced communication.

Context window — A 1M-token window on Opus 4.8 (200K on Sonnet/Haiku) handles entire books, lengthy contracts and large codebases in one pass.

Enterprise traction — Anthropic reports 300,000+ business customers, and Claude is widely adopted in regulated industries that value its safety posture. Notable customers include Salesforce, Netflix, Snowflake, Cognizant and Palo Alto Networks.

Constitutional AI — Anthropic’s safety approach creates more predictable guardrails, valued for compliance, and Claude refuses harmful requests more consistently than many competitors.

Artifacts — Real-time code visualisation and interactive components remain more mature than most Canvas-style equivalents.

Limitations

No image generation — Claude cannot create images. No DALL-E, no Imagen, no Midjourney equivalent. If you need image generation, ChatGPT or Gemini are better choices.

No video generation — Unlike Gemini (Veo) or ChatGPT (Sora), Claude has no video creation capabilities.

Voice is text-to-speech — Despite 2026’s desktop dictation and multilingual mobile voice, the underlying voice is still text-to-speech rather than fully native audio, so it isn’t as fluid as ChatGPT’s real-time voice.

Rate limits — A long-standing complaint. Anthropic doubled 5-hour limits and lifted weekly limits in 2026, but Max plans still carry weekly caps and heavy users can hit them. Power users should check the current limits before relying on Claude all day.

Consumer reach — At roughly 30 million consumer monthly active users, Claude is far smaller than ChatGPT in the consumer market; brand recognition remains a challenge despite strong capabilities.

Cowork and Design are paid — The two flagship 2026 features need a paid plan, and Claude Design is metered separately, so the free experience is still chat-centric. Cowork is no longer desktop-only — since 7 July 2026 it also runs on web, iOS and Android (beta, from Max first) — but it remains a paid feature.

Browser extension still gated — Claude in Chrome is on Max and Team, with Pro rolling out from 12 August 2026; it does not yet support other Chromium browsers or mobile.

Output is watermarked — since August 2026 Claude’s text carries an invisible statistical mark and generated files carry C2PA metadata (see “Watermarking and provenance” below). Anthropic frames it as transparency; users who prefer unmarked output have no opt-out.

Who should use Claude?

Great for:

• Software developers — Claude leads coding benchmarks, and Claude Code (now with its own desktop app) is genuinely transformative
• Knowledge workers — Cowork can take a goal and return finished work on your own files, including scheduled, recurring tasks
• Writers and content creators — Natural, nuanced writing that beats most rivals on long-form work
• Anyone making decks or quick visuals — Claude Design turns a prompt into editable layouts, prototypes and slides
• Researchers and long-document work — A 1M-token context handles entire papers, books and contracts
• Safety-conscious enterprises — Predictable guardrails and strong compliance posture

Look elsewhere if:

• You need image or video generation — ChatGPT (DALL-E/Sora) or Gemini (Imagen/Veo) are necessary; Claude generates designs and HTML, not photos or video
• You want the most generous free tier — ChatGPT and Gemini free tiers allow more casual use; Cowork and Design are paid-only
• Voice-first interaction — ChatGPT’s real-time voice is more natural
• You want all-in-one multimedia — ChatGPT offers text, images, voice and video in one interface

Historical timeline

Privacy and data handling

Understanding how Anthropic handles your data is crucial—particularly after the August 2025 privacy policy changes.

Watermarking and provenance

Since August 2026, Claude’s output itself carries provenance signals — a first among the major frontier labs (Anthropic):

• Text: models launched on or after 2 August 2026 weave an imperceptible statistical watermark into generated text. Anthropic says it doesn’t change the meaning, quality or readability of responses, travels with copy-paste, and “may persist through some editing.” Older models are being transitioned.
• Files: supported generated file types (.svg, .png, .jpg) carry signed C2PA provenance metadata.
• Coverage: the Claude API, the Claude apps, Claude Code, Cowork and Claude Tag — worldwide, not just in the EU, whose AI Act Article 50 (enforceable 2 August 2026) prompted the move (TechCrunch).
• Detection: a detection mechanism is promised in forthcoming technical documentation; there is no user opt-out.
• The caveat Anthropic itself makes: a detected mark shows content was processed by Claude — proofreading and translation count — not that Claude authored it, and an absent mark proves nothing. Short passages and heavily edited text may carry no detectable signal.

The move split users: some object that routine AI assistance at work or school now leaves a trace, while others argue unmarked AI text is the real problem (TechCrunch). For how token watermarking actually works and what it can and cannot prove, see our AI watermarking guide.

Training data policy

Important change: As of August 2025, consumer plans (Free, Pro, Max) default to ON for training data usage. This was criticised as a “dark pattern” since the toggle is pre-enabled.

To opt out: Settings → Privacy → Toggle “Help improve Claude” OFF. This only protects future conversations—data already submitted cannot be removed from training.

Incognito mode

Incognito mode provides privacy protection:

• Conversations aren’t saved to your history
• Not used for model training regardless of settings
• Available to all users via the ghost icon

Data retention

Security certifications

Anthropic holds:

• SOC 2 Type II — Security, availability, processing integrity
• ISO 27001:2022 — Information security management
• ISO/IEC 42001:2023 — AI Management Systems (first major AI company to achieve)

GDPR and international compliance

Anthropic maintains GDPR alignment with:

• Data Processing Agreements available
• Standard Contractual Clauses for EU data transfers
• Data subject rights (access, deletion, rectification)

Enterprise features

• SSO — SAML-based single sign-on
• SCIM provisioning — Automated user management
• Audit logs — 180 days of activity tracking
• Custom data retention — Configurable policies
• VPC isolation — Available via AWS Bedrock or Google Vertex AI
• HIPAA — Configurable with Business Associate Agreement

Visit Anthropic’s Trust Center for compliance documentation.

Claude vs alternatives

Quick comparison

Where Claude wins

• Coding — Leads SWE-bench Verified and the harder SWE-bench Pro with Opus 4.8
• Autonomous work — Cowork completes tasks on your own files; no rival app matches it for non-developers
• Writing quality — Consistently praised as more natural than competitors
• Context window — 1M tokens on Opus 4.8 for whole-codebase and whole-document work
• Safety posture — Predictable guardrails trusted by regulated industries

Where competitors win

• ChatGPT — Image and video generation, real-time voice, persistent memory on the free tier, custom GPTs, larger consumer ecosystem
• Gemini — Deep Google Workspace integration, video generation (Veo), and a generous free tier

Getting started

Web

1. Go to claude.ai
2. Sign up with email or Google account
3. Start chatting—no setup required

Mobile

• iOS: Download from App Store (requires iOS 17.0+)
• Android: Download from Google Play (requires Android 8.0+)

Desktop

Download from claude.ai/download:

• macOS: Apple Silicon and Intel supported
• Windows: Windows 10 or 11
• Note: No Linux version available

Tips for new users

Use Projects for organised work: Create separate Projects for different tasks—each maintains its own context and can have custom instructions.

Upload reference documents: Claude’s large context (up to 1M tokens on Opus 4.8) excels at processing entire documents. Upload style guides, technical specs or research papers for better outputs.

Try extended thinking for complex tasks: Enable extended thinking when tackling difficult coding problems or analysis. Watch Claude’s reasoning to understand its approach.

Explore Artifacts: Ask Claude to create interactive visualisations, code previews, or documents. Artifacts make reviewing and iterating faster.

Set up MCP connectors: If you use Google Workspace, GitHub, or Slack, connect them via Settings to let Claude access your data directly.

Frequently asked questions

Is Claude free?

Yes. The free tier includes Claude Sonnet 5 and Haiku 4.5, web search, file uploads, voice mode and 5 Projects. Rate limits are stricter than paid tiers (~30-40 messages per 5-hour window), and the new Cowork and Claude Design tools are paid-only.

Is Claude Pro worth $20/month?

For most regular users, yes. Pro unlocks Opus 4.8, Memory, Claude Code, and — new in 2026 — the Cowork desktop agent and Claude Design, plus 5x the free usage. It is the sweet spot unless you regularly hit limits.

What is Claude Cowork?

Cowork is Claude’s agent that completes multi-step work on your own files and apps — give it a goal and it returns a finished deliverable, and you can schedule recurring tasks like a daily email triage. It started on the desktop app and, since 7 July 2026, also runs on web, iOS and Android (beta), with tasks executing in the cloud so they continue while your laptop is closed. It is available on all paid plans.

What is Claude Design?

Claude Design is a tool inside the app that turns a prompt into editable visuals — landing pages, app screens, decks, one-pagers — as live HTML you refine with voice, comments or sliders, then export to Canva, PDF, PPTX or HTML. It is in research preview on paid plans and metered separately.

Can Claude generate images or video?

No. Claude does not create photos or video. Claude Design generates editable layouts and HTML, not raster images — for images or video use ChatGPT (DALL-E/Sora) or Gemini (Imagen/Veo).

Does Claude remember previous conversations?

Yes, on paid plans. Memory lets Claude remember preferences and context across sessions; free users must provide context each conversation.

What file types can I upload?

PDF, DOCX, XLSX, CSV, TXT, HTML, PNG, JPEG, GIF, WebP and various code files. Up to 30MB per file, 20 files per chat.

Does Claude watermark its output?

Yes. Since August 2026, Claude models launched on or after 2 August embed an invisible statistical watermark in generated text, and supported generated files (.svg, .png, .jpg) carry signed C2PA metadata — across the apps, API, Claude Code, Cowork and Claude Tag, worldwide, with no opt-out. Anthropic stresses that a detected mark means content was processed by Claude, not necessarily written by it, and that short or heavily edited text may carry no detectable mark. See What is AI watermarking? for the full picture.

How does Claude compare to ChatGPT for coding?

Claude leads on independent coding benchmarks: Opus 4.8 scores 88.6% on SWE-bench Verified and 69.2% on the harder SWE-bench Pro, ahead of GPT-5.5. Claude Code is also a standout developer tool. See Claude vs ChatGPT for the full head-to-head.

Verdict

Claude has become the AI app for serious work. In 2026 it is no longer just a chatbot: Opus 4.8 leads independent coding benchmarks, while Cowork turns Claude into a desktop agent that does multi-step work on your files and Claude Design generates editable visuals and decks. For developers, writers, researchers and knowledge workers who want quality over novelty, Claude delivers.

Free tier: Good for light use and evaluation — Sonnet 5 is genuinely capable — but rate limits and the lack of Cowork and Design make it chat-only.

Pro ($20/month): The sweet spot for most people. Opus 4.8, Memory, Claude Code, Cowork and Claude Design make it competitive with ChatGPT Plus despite no image or video generation.

Max ($100-200/month): For users who lean on Claude all day and want the largest Cowork, Design and usage allowances.

Claude isn’t the best at everything — ChatGPT offers broader multimedia and a more generous free tier, and Gemini integrates deeply with Google. But for coding, writing, analysis and now autonomous knowledge work, Claude leads, which is why Anthropic counts 300,000+ business customers and is the most valuable AI startup in the world.

Official links