Phase 2: Thinking & Reasoning Techniques · 30 min · ChatGPT, Claude, or Gemini

Reasoning Models vs. Standard Chat Models

Different models need different prompts. Knowing which kind you're talking to is the first step.

Hiring signal: What changes when you use this: you adjust your prompting style based on the model you're using instead of using the same approach everywhere — which means better output from every tool.

What you will learn

Tell reasoning models apart from standard chat models
Understand how prompting changes for reasoning models
Know when to use a reasoning model vs a standard model
Adjust prompting style based on the model type

The Problem

Not all AI models are the same. Some are designed to chat — fast, conversational, good at most things. Others are designed to reason — slower, deeper, better at complex problems but worse at simple tasks. Most people don't know which they're using, and they prompt them the same way. That's like driving a sports car and a pickup truck the same way — both will get you somewhere, but neither will perform well.

The Two Categories

Standard Chat Models

These are the models most people use most of the time. They're fast, conversational, and good at a wide range of tasks.

Examples: GPT-4o, GPT-4.1, Claude Sonnet, Claude Opus (without extended thinking), Gemini Pro, Gemini Flash

Best for: Writing, editing, brainstorming, simple analysis, formatting, translation, code generation, most everyday tasks

How to prompt: All the techniques from Phase 1 apply. Be specific, give context, use few-shot, ask for formats. Chain-of-thought ("think step by step") can help on complex tasks.

Reasoning Models

These models are designed to think deeply before answering. They spend more time (and compute) reasoning through the problem internally, then give you the answer.

Examples: OpenAI o1, o3, o4-mini; Claude with extended thinking enabled; Gemini in reasoning mode (Gemini 2.5 Pro with reasoning)

Best for: Complex multi-step problems, math, logic, scientific reasoning, strategic planning, debugging difficult code, problems where getting it right matters more than getting it fast

How to prompt differently:

Skip chain-of-thought — the model already reasons internally. "Think step by step" is redundant and can hurt.
Be clear and complete — state the problem fully. The model will figure out the reasoning path.
Don't over-constrain — reasoning models work best when you give them the problem and let them solve it, rather than prescribing how to solve it.
Be patient — they take longer to respond. That's the reasoning happening.
Don't ask for speed — if you need a fast answer to a simple question, use a standard model instead.

Reasoning models are not "better" — they're different
A reasoning model is not an upgrade from a standard model. It's a different tool for different jobs. For writing an email, a reasoning model is slower and often worse than a standard model. For analyzing a complex business problem with 10 variables, a reasoning model is dramatically better. Choose based on the task, not the model's reputation.

Unlock the full lesson

You've read the first 2 sections. The rest of this lesson covers How to Tell Which You're Using, When to Use Which, Your Practice Rep — plus a hands-on lab, quiz, and project artifact.

Create a free account to unlock Phase 0 and Phase 1 of every course — no credit card.