Do you ever wonder if there's more to life than being really, really, really, ridiculously good-looking?
Insight: Paid AI (“Thinking Models”) is where the magic of AI actually is.
So what? Pay for it! The ROI on $20/mo is unmatched.
I hear this a lot: "I tried ChatGPT and it wasn't that impressive."
And I get it. The free version is underwhelming. It gives you surface-level answers. It hallucinates. It sounds confident about things it's making up. You walk away thinking, "What's the big deal?"
That's because the free models are like Zoolander. Remember him? Ben Stiller's model caricature who's gorgeous but can barely string a thought together? That's free AI. It looks good. It sounds confident. It has absolutely some idea what it’s talking about.
The paid version? That unlocks models who think.
Thinking Models — Beauty and the brains!
That's right — thinking models. These are the AI models usually locked behind the paywall that actually transform how you get work done. Claude's "Extended Thinking," ChatGPT's "Thinking," Gemini's "Pro" — they don't just react. They pause. They reason through problems before answering. They catch their own mistakes. They consider angles you didn't even ask about.
Free models give you the first thing that comes to mind. Paid models give you the thorough thing, after actually thinking it through. The difference is not subtle.
I pay for all of Claude Pro, Gemini Advanced, and ChatGPT Plus, and I exclusively use the thinking models—not “even though”’ they take more time to respond, but because they take more time to respond. They’re so integral to my day-to-day work that I completely forget that most people still only know about Zoolander.
Here's why you should pay for at least one.
Where it matters most
Thinking models shine when the stakes are real (in other words, your job): complex analysis where you need to prioritize across competing options, strategy questions where the answer isn't obvious, writing that matters like board memos and grant narratives, and — maybe most valuably — debugging your own thinking. Hand it your logic and ask "where am I wrong?" You'll be surprised how often it finds something.
For recipes, free is fine. For anything requiring careful judgment (which is quickly becoming the single most important type of work), you want the model that thinks.
Here’s a specific example. I asked a challenging question to various free (Zoolander) and paid (Thinker) models.
Should we consolidate our paratransit and demand-response into one commingled service? Respond in 15 words or less.
And here’s what came back:
Free (Zoolander) Models | Paid (Thinker) Models | |
|---|---|---|
Depends on your goals, but usually yes—commingling improves efficiency and rider experience. | It depends on your ridership profiles, funding restrictions, and union agreements. Let's dig into specifics. Opus 4.6, Extended Thinking | |
Yes, if zoning, ADA needs, scheduling logic, and funding streams align to support unified operations. 5.1-Instant | Maybe—if ADA eligibility, capacity, geography, and funding align; pilot with safeguards, monitor performance, iterate. 5.2-Thinking | |
Commingling can boost efficiency and lower costs, but ensure you maintain strict ADA compliance. 3-Fast | Yes, for efficiency, if your scheduling software and ADA compliance protocols are robust. 3-Pro | |
Avg Response Time | 5 seconds | 20 seconds |
Even in just 15 words, you can see the difference. The free models stayed surface level: yes, you should probably commingle because of generic efficiency. The thinking models surfaced nuances: union, funding, ridership profiles, software, compliance considerations, all with a tendency to refrain from a recommendation without more information.
It's the difference between asking an intern, and asking a senior strategist with 20 years of experience.
The math
$20/month. $240/year.
If a thinking model saves you four hours a month (and it will) that's 48 hours a year. If you value your time at any reasonable rate, you're getting 10x+ return on that investment. It’s a no-brainer (that’s a terrible double-entendre, in my experience it actually makes you use your brain more).
But here's what's harder to measure: you'll attempt things you wouldn't have before. The grant you didn't have time to research. The analysis you kept putting off. The strategy doc you couldn't structure. Paid AI doesn't just save time. It expands what you're willing to try.
Soon, I will be sharing the stories of real transit leaders experiencing this impact first-hand.
Start here
Pick one. Today, I personally recommend Claude Pro for professional work — you'll notice the difference. Pay for one month. Use the thinking model on something that actually matters to you (change the model drop-down from Sonnet 4.5 to Opus 4.6, then toggle on “Extended Thinking”).
Note: I say “Today” because models change quickly. Six months ago, I’d have recommended Gemini. Twelve months ago, ChatGPT.
I consider thinking models so important to the new way of working that I include three free months of Claude Pro for every participant in my Launchpad workshops. That's how much I believe the "aha" moment matters.
If it doesn't change how you work, cancel. But I don't think you'll cancel.
PS – I'm chuckling right now because I know you can't get those models out of your head.
TransitGPT is where I share practical AI insights for transit leaders. Subscribe if this resonates or forward to a colleague who needs it. And hit reply — I read everything.
Go shatter your glass ceiling.