Every big model release brings the same tide of cancellations, resubscriptions, and renewals.

One month everyone's switching to Claude, the next they're crawling back to ChatGPT.

Claude has usually won these swings but this time feels different.

Since Sol dropped, my feeds are full of people switching to GPT seriously, after a long time on Claude.

Part of it is money: Sol sits inside the ChatGPT plan you already pay for, while Fable 5 stays in Claude's paid plans only until a deadline Anthropic keeps moving (currently 19 July), then bills per token at the highest rate they've ever listed.

Part of it is restrictions, Fable's new safety filters flag more harmless requests than before.

And honestly, Anthropic's communication around Fable access has been bizarre enough, the frustration is fair.

But price and vibes aside, the question is if the output good enough to justify the switch?

So I put them head-to-head, and the results… I don’t like them.

Before we dive in, some catchup:

Partner with us

If today's piece left you wanting a proper handle on AI vocabulary, here is the best resource.

85 terms explained in plain English, structured so each section builds on the last.

Quick "test yourself" checkpoints between sections keep what you learn from evaporating, and 15 minutes of scrolling will leave you a lot more fluent than you were this morning.

Prompt:

Output:

Claude's has one big idea, and it's charming.

But it's also leaning very hard on the theme, everything on the page is shouting farmers' market at you, and after a while it feels like a costume.

Sol's page though. Oof.

Elegant serif headline / an illustrated market stall with cool badges / full page of sections, and a readme telling you what to fix before going live.

It reads like a studio built it for a client. The funny part is the reputation says Sol over-decorates, and it used just 1 animation on the whole page.

Winner: Sol

Prompt:

Output:

Both models did something odd at first. Sol generated a picture of a dashboard, and Claude built a small interactive widget inside the chat.

I had to ask both to give me a website instead.

Sol produced a full financial product. Impressive, and complete overkill.

I asked for a tracker, and I got a neobank.

Claude built a thing I would use.

An add-expense form, editable budgets, a delete button, a donut chart. One detail I loved: the placeholder text in the expense form says "Swiggy dinner". Nobody told it I'm Indian. It just read the room.

This is the one test where the internet's big complaint about Sol, that it over-engineers everything, played out exactly as advertised.

Winner: Fable

Prompt:

Output:

This one is the personality test of the whole experiment.

Sol didn't painted the invitation instead of designing one.

Claude wrote a 300-word design manifesto first, then executed it.

Claude's is the one a design-literate parent frames, but the invite isn't for the parent.

Show both to the 6-year-old and he picks Sol's every single time, because Sol's looks like a party and Claude's looks like a gallery.

Winner: Sol

This is just three prompts, my taste, and a design test, so half of this is subjective by definition.

The benchmarks split the same way the discourse does.

Sol ahead on some, Fable ahead on repo-level coding and overall intelligence scores, and the chart OpenAI used to crown Sol got openly mocked on Reddit as unreadable and gamed.

Also both models needed one nudge on the dashboard test before producing a website, so neither is reading minds yet.

Sol won my design test 2-1, and I'm as surprised as anyone, because I walked in carrying the same assumption most of the internet carries: Claude is the tasteful one.

What I think happened is that reputations move slower than models.

Claude's design instinct is there, but it has hardened into a style, minimal, restrained, cream-coloured, philosophical, and when that style meets a brief that wants joy, it starts looking less like taste and more like a habit.

Sol just read each brief and matched the energy: editorial for the market, chaos for the kid.

Add the money math, and the switching threads stop looking dramatic and start looking rational.

Until next time,
Vaibhav 🤝🏻

If you read till here, you might find this interesting

#AD 1

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

Learn more

#AD 2

When HR and IT don't talk, people fall through the cracks. This guide fixes the handoffs that matter most.