Claude: the models, the apps and when to use which

What it is

Claude is an AI assistant made by Anthropic. You type or speak a request, and it drafts, summarises, analyses files or answers questions. Behind it sits a family of models (the trained systems that write the answers) in different sizes.

Think of a team with a quick assistant, an experienced colleague and two senior specialists. You would not give every job to the most senior person.

The table shows the current names, as of November 2026; check the maker's page, because names change often.

Current models decide how much to "think", step by step, before answering. Anthropic calls this adaptive thinking (see guide 03).

Why it matters

• Claude runs on the website, desktop apps for Mac and Windows, a mobile app, and inside tools you may already use.
• Add-ins work inside Excel, PowerPoint and Word, with Outlook in beta. A Google Workspace add-on is also in beta.
• Claude in Chrome is a browser extension that can read a page and take actions such as clicking and filling forms, so watch every action (see guide 19).
• Cowork handles longer, multi-step tasks such as research briefs, documents and file organising. Claude Code is the tool for software work.
• Plans differ on data. On personal plans (Free, Pro, Max, Ultra) you choose in privacy settings whether your chats help improve Claude. If you allow it, Anthropic keeps that data longer (up to five years, it states). On business plans (Team, Enterprise), Anthropic states it does not train on your content by default.
• Like any assistant, Claude can state wrong facts confidently, and features vary by plan; research results and actions it takes still need a person to check.

How to do it

1. Check which Claude plan and features your organisation has approved, and use your work account, not a personal one (see guide 10).
2. Start with the balanced tier for everyday work; move down for simple, repetitive jobs and up when answers miss important points.
3. Create a project (a workspace holding files and standing instructions) for repeat work, such as monthly reporting.
4. Use the research mode for questions that need many sources (see guide 08), and open the key citations.
5. Connect approved tools through connectors (links that let Claude read your email, calendar or files) only with your administrator's permission (see guide 13 and guide 40).
6. Review every draft, figure and action before it leaves your hands. You are the final check.

Example

Before: pasting a long supplier contract into a chat and asking "anything bad in this?"

After: inside an approved work project, with made-up or cleared data:

You are helping me review a supplier agreement. Read the
attached contract and list:
1. Payment terms, notice periods and renewal dates.
2. Any clause that differs from our standard terms (attached).
3. Questions I should ask the supplier.
Quote the clause number for every point. If unsure, say so.

You then open each quoted clause and confirm it says what Claude reports. The decision stays with you and, where needed, your legal team.

Common mistakes

• Using a personal account for work. Fix: use the approved business account, where training is off by default.
• Always choosing the largest model. Fix: start balanced; the largest models are slower and use more of your limit.
• Letting the browser extension act unwatched. Fix: approve each step and stay on the page.
• Trusting summaries of files without opening them. Fix: ask for quotes or page numbers and spot-check them.
• Assuming features are on every plan. Fix: research, some models and some add-ins depend on your plan; ask IT.

Try it this week

1. Open Claude in your approved work account (2 minutes).
2. Create a project called "Practice", add a made-up two-page document and a one-line instruction about tone (5 minutes).
3. Ask the same summary question with the fast and the balanced model, and compare (5 minutes).
4. Write one line for your team: which model for which job (3 minutes).

Going further

• Memory lets Claude carry useful facts between chats; review and edit what it remembers in settings.
• Voice mode lets you talk to Claude; useful for thinking aloud, less so for precise wording.
• Artifacts are documents, slides or small tools Claude builds beside the chat; check them like any draft.
• Business plan administrators control which models, connectors and features staff use.
• Re-test your team's prompts when a new model arrives (see guide 27).

Key words

• Model: the trained system that reads your request and writes the answer.
• Adaptive thinking: Claude decides how much step-by-step reasoning a task needs.
• Project: a workspace with saved files and instructions for related chats.
• Connector: an approved link that lets Claude read or act in another tool.
• Add-in: a version of Claude that works inside another program, such as Excel.

Sources

Examples and figures in this guide are illustrative. AI tools change often: check what your organisation's approved tools offer today. Use approved tools only, and a person checks and decides.

Prefer to print or share it? Download this guide as a PDF (279 KB), or the complete AI Field Guide (2.5 MB).