Claude 3.7 Sonnet and its Two Thinking Modes
Same assistant, same knowledge... two conversation modes
I've always had a soft spot for Claude. There's something about the way our conversations flow that feels "more human" to me, though I still can't quite put my finger on why.
Now, with Claude 3.7 Sonnet, Anthropic has released an assistant with two thinking modes: Normal and Extended.
If you've read Daniel Kahneman's "Thinking, Fast and Slow", this might sound familiar. Kahneman talks about how our brains have two different decision-making systems: a quick, intuitive one and a slower, more analytical one. What's cool is that Claude now seems to have these systems too.
Just like we intuitively know when to trust our gut versus when to slow down and think step-by-step, Claude can now shift between these two modes... all while being the same assistant underneath. Let me show you how this works.
A Hybrid Reasoning Model
Claude 3.7 Sonnet is what Anthropic calls a "hybrid reasoning model." But what's interesting isn't just that it offers two different modes (which technically we could get by using separate models), but that it's one brain with two different ways of processing information.
This works differently than ChatGPT (and other assistants), where you have to choose between different models like GPT-4o or o1 to get different thinking styles. With Claude 3.7 Sonnet, the same architecture and training powers both thinking modes, it's the same AI at its core, just trained to think in different ways.
In normal mode, Claude gives you quick, to-the-point responses. Switch to extended thinking mode, and suddenly you can watch its reasoning unfold step by step, almost like seeing the thoughts form in real time.
One Brain with Two Ways of Thinking
Picture this: you visit a restaurant where the chef gives you a choice - either enjoy the perfectly plated dish served directly to your table, or step into the kitchen and watch them work their magic with each ingredient. Same chef, same delicious food, but totally different experiences.
Most AI assistants make you switch between different models for different tasks. With OpenAI, I use GPT-4o for everyday conversations, but I jump to o1 or o3-mini when I need deeper analysis. They're basically different AI brains designed to think in different ways.
What makes Claude 3.7 Sonnet special is that it's one brain that can operate in two different ways.
It's not like Claude ducks into a phone booth and emerges as "Super Claude" (sorry for the vintage analogy). It's exactly the same model with all the same knowledge. The difference is whether it shows us its thinking process or just the final answer.
In normal mode, Claude's responses are direct and to the point. Like asking a friend, "Where's the best coffee in town?" and getting "The corner of Main Street" as an immediate answer.
In extended mode, that same friend might say: "Well, to find the best coffee, I need to consider what makes great coffee. There's bean quality, brewing methods, atmosphere... Based on these factors, I'd recommend three places that really stand out for different reasons..."
Why This Difference Matters to Me
It mirrors how our brains actually work: We don't swap out our brains when switching from casual chat to deep problem-solving.
It helps me catch mistakes: Seeing the full reasoning process makes it obvious when something doesn't quite add up.
What Claude is really offering isn't two different tools, but a way to either see its entire thought process or just the final answer. And that changes how I interact with AI.
The Two Thinking Modes
I've been playing around with both modes of Claude 3.7 Sonnet, and I'm starting to get a feel for when each one works best. Here's what I've discovered so far:
Normal Mode for Everyday Use
In normal mode, Claude cuts to the chase. You get clear, direct answers without all the behind-the-scenes mental work.
It's like asking a chef "how much cumin should I add?" and getting a simple "one teaspoon" in response. No lecture on spice profiles or cooking chemistry - just the practical answer you need right now.
This is my go-to mode for 90% of what I do, and it's perfect when I need:
Quick creative sparks: "Hey Claude, give me 5 catchy titles for my blog post about your two thinking modes" 😄
Ready-to-send text: "Draft a friendly email declining this event invitation"
Straight-to-the-point explanations: "Explain inflation in two clear paragraphs"
Natural back-and-forth: When I want a conversation that flows without long-winded explanations getting in the way
Extended Mode: When You Need to See the "How"
This is where things get really cool. Extended mode doesn't just give you Claude's answers but shows you exactly how it got there, one step at a time.
It's not just about WHAT Claude thinks, but HOW it arrived there.
Think of our chef analogy again - instead of simply handing you a teaspoon of cumin, the chef invites you into the kitchen. "Come in, let me show you how I balance these flavors, why this spice works with these ingredients, and how I know exactly when it's enough." Same chef, same dish, but you gain insights you'd never get from just eating the final meal.
The best part? Extended mode doesn't just show you the thinking behind the answer. When Claude thinks step by step, it often comes up with different (and usually better) answers than in normal mode. It's like watching someone solve a puzzle - when they slow down and work through each possibility one by one, they spot solutions they might miss when rushing.
This mode becomes invaluable when:
You need rock-solid confidence: Seeing each reasoning step lets you spot any logical flaws or questionable assumptions
The problem is really tricky: When it's the kind of question where three smart friends would give you three totally different answers
You want to get better at problem-solving: By watching Claude's problem-solving approach, you pick up frameworks for tackling similar challenges yourself
Here's what it looks like in action:
A Real-World Example
This example shows how Claude approaches problems in extended mode, revealing its entire thought process.
First I ask a question…
Me:
I've been working for five years as a financial analyst at a large company, with good salary and stability. I was offered a position at a startup, where the initial salary is lower but with stock options and the possibility of accelerated growth. What factors should I evaluate in depth before making a decision?Then Claude starts thinking…
Claude's thinking process (took 7 seconds):
The person is asking me to help them evaluate factors they should consider before deciding between staying at their large company where they have a good salary and stability, or accepting an offer at a startup with lower initial salary but stock options and potential for accelerated growth.
This is a career decision that involves weighing various factors related to financial implications, career growth, risk tolerance, and personal preferences. Let me think about what factors they should consider in depth:
Financial considerations:
- Current compensation gap - how much lower is the startup salary and can they afford this reduction?
- Valuation of stock options - what percentage of the company do they represent? What's the current valuation? What's the vesting schedule?
- Benefits comparison - health insurance, retirement plans, other perks might differ significantly
- Potential financial upside - realistic assessment of stock option value if the startup succeeds
- Long-term financial planning - impact on savings, debt repayment, major life purchases
Career growth and development:
- Learning opportunities at each company
- Skills development potential
- Career trajectory and advancement speed
- Responsibility level and role scope
- Résumé building/career marketability
- Network expansion opportunities
Startup-specific considerations:
- Funding status and runway (how much money they have, how long it will last)
- Business model viability and market opportunity
- Leadership team experience and track record
- Growth metrics and traction so far
- Competitive landscape
- Exit strategy possibilities (acquisition, IPO)
Stability and risk factors:
- Job security differences
- Startup survival statistics (most fail)
- Personal financial safety net
- Industry and economic conditions
- Personal risk tolerance
Work-life factors:
- Work culture differences
- Hours/flexibility expectations
- Commute/location considerations
- Work-life balance
- Alignment with personal values and interests
- Team dynamics and potential colleagues
Personal considerations:
- Career goals and how each option helps achieve them
- Life stage and personal circumstances (family obligations, etc.)
- Passion for the startup's mission vs. current role
- Long-term career direction
I'll put these factors together in a comprehensive but concise response that helps the person evaluate this important career decision.After that Claude is ready to respond…
Claude response:
When considering a move from your established financial analyst role to a startup opportunity, here are the key factors to evaluate:
Financial Considerations
Quantify the total compensation gap (including benefits)
Assess stock option value (percentage ownership, current valuation, vesting schedule)
Determine if you have sufficient savings to manage the salary reduction
Evaluate the startup's funding status and runway
Career Development
- Compare growth trajectories at both companies
- Assess skill development opportunities
- Consider role scope expansion and leadership potential
- Evaluate how each path enhances your market value
Startup Viability
- Research the founding team's track record
- Analyze the business model and market opportunity
- Examine growth metrics and customer traction
- Understand the competitive landscape and exit potential
Personal Factors
- Honestly assess your risk tolerance
- Consider work culture fit and values alignment
- Evaluate life circumstances (family needs, financial obligations)
- Reflect on which opportunity better serves your long-term career vision
Remember that startup compensation is high-risk/high-reward, while larger companies offer stability but potentially slower advancement. The right choice depends on your personal circumstances, career goals, and comfort with uncertainty.I love seeing how Claude organizes its thinking! Did you notice how it first confirms it understands my question? Then it puts together a solid framework before giving a clear, well-structured answer.
This window into its reasoning process shows exactly how it gets from question to answer (something we'd never see in normal mode).
Thinking in Action
Think of extended mode as the "behind-the-scenes" bonus features on a DVD (another vintage reference 😊). Most of us just watch the movie and enjoy it, but sometimes we're curious about how the magic happened behind the camera.
Normal mode is my go-to for almost everything:
Natural conversations ← (this is where Claude truly shines for me)
Everyday questions that need quick answers
Creative writing and brainstorming sessions
Pretty much anything I need to handle efficiently
When to Fire Up a New Chat in Extended Mode
There are specific situations where it's worth starting a fresh chat in extended mode Important note: you can't go from one mode to another in mid-conversation, but you can go from normal to extended mode selecting the “Retry” action in a response, Once you change to extended mode you cannot go back to normal model in that conversation (at least I haven’t found out how)
Extended mode really pays off when:
You're dealing with high-stakes information: When you're working on something that really matters or trying to make decisions based on lots of data, seeing Claude's reasoning helps you spot any flaws in the logic or conclusions.
You're tackling complex subjects: It's like having a master teacher who doesn't just hand you the answer but walks you through their entire problem-solving approach, helping you build your own understanding.
Something feels off about an answer: We've all had that moment when an answer seems reasonable but something about it just doesn't feel right. When that happens, try the same question in extended mode.
What's Working For Me
Here's what I've been doing that seems to work pretty well:
Default to normal mode for almost everything (saves time and flows better)
Open a new extended mode chat when you need deeper insight or have doubts
Ask the same question (sometimes adding a bit more context)
Compare both answers to see the difference in depth and reasoning
Pro tip: Remember you can't switch modes within a conversation - you must start a completely fresh chat with extended mode selected from the beginning.
Bonus Track: Claude Code
When Anthropic launched Claude 3.7 Sonnet, they also introduced something special for the developers among us: Claude Code.
This new tool lets coders give tasks to Claude straight from their command line - basically bringing Claude's smarts right into how they work. If you want to see how it works, there's a great video that shows it all.
Where We're Heading
I really like how Anthropic has approached Claude 3.7 Sonnet, giving us two thinking modes in one model instead of making us choose between different models like all other AI companies do. This all-in-one approach just feels more natural and intuitive.
My wish for the future? Being able to seamlessly toggle between modes mid-conversation. Imagine chatting with Claude and simply saying, "Hold on - can you show me your thinking on that?" just like you might ask a friend to explain their reasoning. That kind of flexibility would be a game-changer!
See you around!
G
PS: Phone booths? DVDs? I guess my vintage tech references are giving away my age...
Hey! I'm Germán, and I write about AI in both English and Spanish. This article was first published in Spanish in my newsletter AprendiendoIA, and I've adapted it for my English-speaking friends at My AI Journey. My mission is simple: helping you understand and leverage AI, regardless of your technical background or preferred language. See you in the next one!