Pick the right tool and model

01What this module is about

Every AI tool gives you choices: which model to use, and how hard it should think. Most people never touch these settings. Some always pick the biggest one. Both waste time.

This module gives you a simple rule for choosing, shows you where the settings are, and teaches you to test a tool on your own work instead of trusting rankings.

By the end, you'll know which setting to start with for any task, and when it's worth changing.

02What a model is

A model is the engine behind the chat. Each company makes several, from small and fast to large and slow.

• Faster models answer in seconds. They're fine for everyday tasks: rewording, short emails, quick questions.
• Stronger models take longer and think things through. They're better at hard reasoning, long documents, careful comparisons and code.
• Thinking is a setting most tools now have. The AI works through the problem before it answers. More thinking usually means a better answer to a hard question, but a slower one.

Stronger models and more thinking also use up your plan's allowance faster. Google says this directly about Gemini. That's why the biggest setting isn't always the best choice.

03The rule: start with the default, step up when it falls short

1. Start with the default, or Auto. Most tools now pick for you, and they pick well for everyday work.
2. If the answer falls short, turn up the thinking first. Try the same request again with more thinking.
3. If it still falls short, switch to a stronger model. Anthropic gives the same order for Claude: raise the effort first, then switch to a more capable model.
4. Go back to the default for the next ordinary task.

Signs you should step up: the answer misses steps, mixes up details from a long document, or gets a calculation wrong.

04Where the settings are

The settings sit next to the message box in every tool. Names change often, so learn where to look, not the names. This is what they showed on 11 November 2026:

• Claude. Click the model name next to the send button. You'll see the models, plus Effort, which sets how much it thinks. The models are Haiku (fastest), Sonnet, Opus and Fable (strongest). Free accounts get Sonnet and Haiku. On the Pro plan, Fable costs extra. It's paid for with usage credits, which Module 8 covers.
• ChatGPT. Click the button next to the microphone. It opens Thinking effort, a slider that starts at Instant. Slide it right for more thinking. Paid plans can also choose the model.
• Gemini. Click Gemini Auto in the message box. Auto picks for you. Thinking level lets you choose Low, Medium or High.
• Grok. Click Fast in the message box. The choices include Auto, Fast, Expert and Heavy. Which ones you get depends on your plan.
• Claude Code. It uses the same Claude models as Claude.

05Which tool for what

All five tools can do most everyday tasks well. You don't need all of them. Use the one you know, and reach for another when it has something yours doesn't.

• For today's news and live events, use a tool that searches the web. Claude, ChatGPT and Gemini all can. Grok can also search public posts on X, which helps with breaking news. Always open the sources it gives you, because fresh news is often wrong.
• For work on files on your computer, Claude Code can read and change them directly. It needs a paid Claude plan.
• For a second opinion, use a different company's tool. Two tools from the same company tend to share the same blind spots.

06Text to copy

When an answer matters, get a second opinion. Paste the first answer into a different company's tool with this:

Here's an answer from another AI. Check it carefully.
List anything that's wrong, missing or not backed up.
Don't rewrite it. Just list the problems, most serious first.

[paste the answer]

To test a tool on your own work, keep one task you can reuse. Save it in your project folder as TEST-TASK.md:

[A real task from your project that has a right answer you already know.
For example: "Using the Reference file, list every walk in March with its date, distance and meeting point."]

08Common problems

• The answer is slow and no better. You don't need the extra thinking for this task. Go back to the default.
• You've hit a usage limit. Stronger models use up your allowance faster. Switch down, or wait for the limit to reset.
• The setting you want is greyed out. It's probably on a higher plan.
• Two tools disagree. Don't pick the one you like. Check the facts yourself. Module 7 shows you how.

09Keeping it current

• Models change every few months. When your tool announces a new one, run your test task on it before you rely on it.
• Don't trust rankings to choose for you. Rankings test general puzzles, not your project. Your test task tells you more.
• Have a backup tool. If your main tool is down or out of allowance, you can carry on in another, using the fallback from Module 1.