Claude

Rating reflects Anthropic's documented plans and capabilities as of September 2026 plus our own sessions. Some vendor links on this site are affiliate links; they never affect a rating. See our disclosure and method.

What it does

Claude is Anthropic's chat assistant, and in 2026 it is closer to a workbench than a chatbot. The web and desktop apps split into three modes: Chat, for the ordinary back-and-forth; Cowork, which hands a task to Claude to run in the cloud while you do something else; and Code, which is Claude Code, the terminal agent that reads your repository, edits files, runs tests and iterates until the job is done. Around those sit web search, file uploads, artifacts (a side panel where documents, code and small apps appear and can be edited), Projects for keeping related files and instructions together, task scheduling, a Chrome extension, a Microsoft 365 extension, and connectors to tools like Slack, Google Drive and Notion. Anthropic's pricing page also lists Claude Design, Slides, Docs and Science as included apps on paid plans.

The model lineup changed three times this year. Claude Sonnet 5 arrived on June 30 and is the default on Free and Pro. Claude Opus 5 followed on July 24 and is the default on Max and the strongest model on Pro. Claude Fable 5.1, Anthropic's top model, became generally available in September; the pricing page lists it on Pro with usage credits rather than as part of the flat fee, and independent trackers report that Max includes it inside the weekly limit. Haiku 4.5 is there for quick, cheap work. In practice most people will live on Sonnet 5 and reach for Opus 5 when a draft or a bug needs more care.

Memory got a proper overhaul on August 25. Claude now adds topics to memory as you chat instead of summarizing afterward, the same memory is shared between chat and Cowork, and everything it has stored sits in a list under Settings, Memory, where you can read, edit or delete each file. Sensitive topics such as health, religion and politics are not stored unless you turn that on. Search across past chats is a paid-plan feature; memory itself is listed on the free plan.

Where it wins

• Prose. Short emails, long reports, tone changes and edits come back sounding like a person wrote them. In our prompt-pack testing Claude needed the least editing of the three flagship assistants, and reviewers through 2026 keep reaching the same conclusion.
• Long documents. Upload a contract, a thesis or a 300-page PDF and ask questions across the whole thing. Anthropic's current models run a one-million-token context window on the API; the apps handle book-length material without chunking.
• Claude Code on the $20 plan. Pro includes the terminal agent that most developers would otherwise pay for by the token. It plans, edits, runs tests and fixes failures on its own. It shares your plan's usage budget, so heavy sessions eat into chat.
• Memory you can read. Every remembered fact is a short editable file. Fix your company's old name once and every later conversation gets it right. It is the most transparent memory system in the suit.
• Working with your files. The desktop app, Cowork and the Microsoft 365 and Chrome extensions make Claude the strongest option for people whose work lives in documents rather than in Google's apps. BGR's July review called the desktop app the point where "the product really comes into its own".

Where it loses

• Usage limits. Paid usage runs in five-hour sessions with a weekly cap on Max. Pro promises "at least five times" the free allowance and never a number; one reviewer hit a ceiling around 45 messages in five hours during heavy work. ChatGPT's free tier is more generous for plain chat.
• No image generation. Claude can read images and build charts and diagrams, but it does not make pictures. ChatGPT and Gemini both do.
• The training default. Since the 2025 consumer terms change, the Help improve our AI models toggle is on for Free, Pro and Max unless you turn it off, and leaving it on extends retention of your chats to five years. Business plans and the API are excluded. Two clicks fix it; you still have to make them.
• Thin bundle at $20. Pro buys models, Claude Code and Projects. Google's AI Pro is the same money with 5 TB of storage, Gmail and Docs integration and YouTube Premium Lite.
• The top model is metered. Fable 5.1 is the model in the headlines, and on Pro you pay for it by usage credit. If you want it as part of a flat fee you are looking at Max.

Plans and pricing, September 2026

Team seats are $20 a month (standard) or $100 (premium, with 5x the usage) on annual billing, $25 and $125 monthly, with a two-seat minimum and no training on company content by default. Enterprise is $20 a seat plus usage at API rates, quoted on request. Max cannot be billed annually.

Score breakdown

Headline rating weights Capability 30%, Ease of use 20%, Value 20%, Trust 15% and Ecosystem 15%, the same weights we use for every card.

Alternatives

• ChatGPT if you want the most generous free tier, image generation and the widest set of extras.
• Gemini if you live in Gmail and Docs and want storage and Workspace features bundled into the $20.
• Perplexity if most of your questions are research questions and you want the sources first.

Quick answers

Is Claude free?

Yes. The free plan runs Claude Sonnet 5 with a usage limit that resets every five hours, plus web search, file uploads, artifacts and memory. It does not include Claude Code, Projects or the bigger models. Pro is $20 a month, or $17 a month billed annually.

Does Claude train on my conversations?

By default, yes, on Free, Pro and Max. Since the 2025 consumer terms change the Help improve our AI models setting is on unless you switch it off under Settings, Privacy, and leaving it on extends retention to five years. Team, Enterprise and API use are not trained on.

Is Claude Pro or Max worth it?

Pro is worth $20 if you write or code most days: you get Opus 5, Claude Code, Projects and at least five times the free usage. Max at $100 or $200 only makes sense if you run Claude Code for hours or want Fable 5.1 without paying per use.

More cards

• 
  Tool cards

  ChatGPT

  The model-first assistant, rated and priced.

• 
  Tool cards

  Cursor

  The agent-first code editor, if Claude Code is not enough.

• 
  Guides

  Best AI chatbot 2026

  Four assistants, one table, one pick per person.