Reasoning models and effort

Why it matters

Modern models can think before they answer. They work through a problem in private, then write the reply. That's why they can now plan an analysis, fix broken code or weigh the options in a brief. But thinking costs tokens, and tokens cost time and money. You can set how hard the model thinks, one task at a time. That setting is one of the biggest levers you have on quality, speed and cost.

What you'll be able to do

• Explain what thinking is and why it improves answers to hard problems
• Set the effort level on a request and choose it for a task
• Read thinking blocks in a response, and ask for readable summaries
• Budget max_tokens and cost for requests that think

Inside this lesson

• 11 short sections
• 4 code examples you can edit and run
• a practice exercise that checks your code
• a 5-question quiz
• 4 review cards for spaced repetition

What it covers

1. What thinking is
2. Adaptive thinking and effort
3. Reading thinking in a response
4. Thinking costs tokens
5. Worked example: what does more effort cost?
6. Keeping thinking blocks in conversations

A taste of the code

body = {
    "model": "claude-opus-5-5",
    "max_tokens": 8000,
    "output_config": {"effort": "low"},   # low, medium, high, xhigh or max
    "messages": [{"role": "user", "content": "Tag this review: 'Arrived chipped.'"}],
}

Try a quiz question

Before this lesson

• Choosing a model