Big Idea

The argument in one line.

Fable 5 thinks meaningfully more layers ahead than Opus 4.8 on complex multi-part builds, but for well-specified simple UI tasks the depth gap closes fast enough that the 2x token cost is hard to justify.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You are already using Claude Code or Claude.ai to build apps and want a data-backed decision on when to upgrade your model.
You are building something with many interdependent moving parts and want to know if a smarter model changes the outcome.
You are on a Claude subscription and need to understand the June 22 pricing cliff before it hits your workflow.

SKIP IF…

You want benchmark tables rather than live-coded demos.
You only build simple static pages or landing sites — the verdict is already clear: Opus handles that fine.

TL;DR

The full version, fast.

Anthropic released Fable 5 and restricted Mythos 5 simultaneously; Fable 5 is Mythos with safety guardrails, making it the new top model above all prior Claude tiers. After 24 hours building four apps side-by-side at high effort, the verdict is nuanced: Fable adds unprompted features, avoids generic design defaults, and creates dramatically richer simulations. But for a landing page or simple booking UI, a more detailed Opus prompt closes the gap at half the cost. The rule: Fable for planning, architecture, and compounding complexity; cheaper models for well-specified execution.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 02:47

01 · Intro + Pricing Context

Fable 5 is trending; free on Claude plans until June 22 then credits-only at 2x Opus cost. Test: four apps, identical prompts, Fable vs Opus.

02:48 – 03:16

02 · Model Landscape

Anthropic released Mythos 5 (restricted) and Fable 5 (Mythos + safety guardrails) simultaneously. Fable sits above all prior Claude tiers.

03:17 – 07:29

03 · Use Case 1: Mobile Flashcard App

CocoQuiz iOS-style app modeled on Quizlet. Fable adds an unsolicited matching game with drag-and-drop and timer. Both look strong; Fable adds unprompted depth.

07:30 – 09:26

04 · Use Case 2: Calendly-Style Booking App

Calendarr web app. Opus defaults to purple SaaS; Fable picks a warm editorial palette. Functionality equivalent. Not worth 2x premium for this complexity level.

09:27 – 11:51

05 · Use Case 3: Voice Dictation Landing Page

VoiceFlow SaaS landing page. Fable produces five sections vs Opus three, better testimonials, avoids purple. Wins on design but prompt engineering closes the gap.

11:52 – 14:47

06 · Use Case 4: SimCity-Style Browser Game

Isometric city simulation stress test. Opus: uniform boxes, box cars, basic traffic. Fable: districts, named citizens with behaviors, moving sun, traffic jams. Clear winner.

14:48 – 16:14

07 · Personal Take + When to Use Fable

Practical decision rule: Opus/Sonnet for simple tasks and well-specified execution; Fable for complex planning, multi-part architecture, and pre-launch security audits.

Atomic Insights

Lines worth screenshotting.

Fable 5 is Mythos 5 with safety classifiers added — the same underlying model, restricted for general release.
Fable 5 costs 2x Opus 4.8 and is only free on existing Claude plans until June 22, 2026.
Setting Fable to max or extended thinking can run a single prompt for hours; high effort is the practical sweet spot.
Fable adds features you did not specify — it reasons about what the brief left out and builds ahead.
The purple SaaS default is a reliable signal that a model is playing it safe rather than reasoning about design.
Fable's advantage is reasoning depth, not speed or token efficiency — the gap widens as problem complexity scales.
For planning and architecture, one Fable session can replace many Opus follow-up correction prompts.
A pre-launch security audit is a high-value, infrequent task that fits Fable's cost profile well.
The hand-off model works: Fable plans the architecture, Sonnet or Opus executes the implementation.
The SimCity demo is the clearest signal: identical prompt, dramatically different simulation depth and detail.

Takeaway

Three cases where Fable 5's reasoning depth actually changes the output.

WHAT TO LEARN

The gap between Fable 5 and Opus 4.8 is real but task-dependent — it shows up clearly on complexity and disappears on simple, well-specified work.

For simple, well-specified tasks like landing pages or basic web UI, a more detailed prompt to Opus achieves comparable results at half the token cost.
Fable 5 reasons about what the brief left out and builds ahead — its advantage is in tasks where anticipating omissions is the hard part.
The clearest trigger for Fable is many interdependent moving parts: CRMs, desktop apps, simulations — anything where one missed dependency cascades into broken states.
Architecture and planning sessions are the highest-leverage use of a frontier model: spend one Fable session mapping structure, then hand off execution to a cheaper model.
A pre-launch security audit — checking for exposed API keys, insecure DB configs, access control gaps — is a high-value, low-frequency task that justifies a frontier model's cost premium.

Glossary

Terms worth knowing.

Fable 5: Anthropic's publicly available frontier model released June 2026, built on the same base as Mythos 5 but with safety classifiers that block high-risk outputs.
Mythos 5: Anthropic's unrestricted super-intelligent model released simultaneously with Fable 5; available only to a small group of vetted users due to its capabilities.
Vibe coding: Building software by describing intent in natural language and letting an AI model generate the implementation, iterating by feel rather than writing code manually.
High effort mode: A Claude reasoning setting that increases compute and depth without running to the extreme of extended thinking or max mode, which can consume hours on a single prompt.
Credits-only pricing: A billing model where each API call is metered per token on top of a base subscription plan, as opposed to unlimited usage included in the monthly plan.

Resources

Things they pointed at.

03:54productQuizlet ↗

07:32productCalendly ↗

01:37linkEden AI benchmark comparison ↗

00:00linkthebreadcrumb.co ↗

Quotables

Lines you could clip.

01:47

“Fable five is a model built for ambitious and long running work — it's not really designed for simple one off tasks.”

Anthropic's own framing delivered clearly; instantly reusable as a model positioning clip.→ TikTok hook↗ Tweet quote

06:05

“I didn't even ask it to do this — it just added a matching game. That's something impressive.”

Concrete proof-of-unprompted-reasoning moment; strong reaction clip.→ IG reel cold open↗ Tweet quote

14:48

“If you're building something simple like a landing page, Opus 4.8 will give you basically the same result. There's no real reason to spend twice as much.”

Honest counter-hype take that earns trust.→ newsletter pull-quote↗ Tweet quote

15:42

“Did I accidentally expose my API keys? Is my database secure? This is the kind of thing Fable is really good at flagging — I use it like a final check before I launch anything.”

Actionable novel use case most viewers would not have thought of.→ TikTok hook↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

analogystory

00:00Cloud just released a brand new super intelligent model called Fable five. It's been trending everywhere and I've been testing it for the past twenty four hours and in this video we're gonna see if it actually lives up to the hype. But instead of just showing you some benchmark, I wanna show you some practical use cases.

00:14I'm gonna show you some of the things that I actually built. Some of it was okay, but but some actually genuinely blew my mind. Now there's one thing to note.

00:21Fable five is now available to use in your monthly cloud subscription, but only until June 22. And after that, it moves to credit only, meaning you'll pay per use on top of your plan.

00:32And it also costs twice as much as Opus. So by the end of this video, you'll know whether this is all worth it, where it actually falls short when you're building apps, and exactly when to use it versus when to stick with cheaper models. But before I show you the apps that I built, I wanna explain what this model actually is because there's a bit of backstory that makes it quite interesting.

00:49Now just about a day ago, Anthropic actually released two models at the same time, Fable five and also Mythos five. If you've following Clot over the past few weeks, you've probably heard the rumors about a superintelligent model that they're working on called Mythos. There was a lot of speculation around it, and Anthropic even came out and said it could be dangerous in the wrong hands.

01:07So instead of releasing Mythos to everyone, they built a safer version called Fable five. So basically the same model underneath, but they added some guardrails that block all the dangerous stuff.

01:18Right? So Fable five is essentially Mythos with some safety layers on. Now if you've been using Claude for a while, wanna show you where Fable five sits compared to the other models.

01:26So you've got Haiku, which is the smallest and the fastest, and then you got Sonant, which is the everyday model. And Opus 4.8 was until recently the most powerful model, but now Fable five sits above all of them. And it's not just the best slot model but right now Fable five beats every model from Gemini, Charge dbt 5.5, and Grok four across pretty much everything.

01:47Capable but it also cost twice as much as Opus, which means that you wanna know when to use it and when not to use it. And this is actually how Anthropic intends for you to use it. So they describe Fable five as a model built for ambitious and long running work, meaning it's not really designed for simple one off task, but it's designed for long running projects.

02:06So the way to think about it is that you wanna use Opus or Sonnet for simple task. We use Fable five when you're doing some advanced planning for an app that has tons of features, something that would normally take you, like, 20 prompts to get through. Give it the full problem, let it run, and it will think through the whole thing on its own.

02:20Alright. So let's get right to the demos, the apps that I was able to build with Fable five. So I had the chance to build four different apps with Fable five, a mobile application, a web application, a SaaS landing page, and also a simulation game to really stress test the model.

02:34And the way that I built these apps is that I ran the same exact prompt and built it using Fable five and also Opus 4.8 just to compare the results. And I'm gonna show you both results so you can tell me which one you think is the best. Now one thing to flag before I show you these demos, when you select the model Fable five here, you can actually select the effort level as well and I actually put it in high effort, not extended thinking or max.

02:57The reason is that if I put it in max or extra high, a single prompt can just run for hours and it's just gonna obliterate your tokens very very quickly. So I found that high effort is kind of the sweet spot. So I kept it at high effort here and I also use the same setting when building the app with Opus.

03:11Now normally when I build any app with AI, I always start by asking it to plan out the features first. That is a habit that I would recommend because it keeps the output focused and it prevents you from getting AI slop. But we're gonna do something different today because the whole point of this test today is to give Fable five just enough information just to see how well it reasons and think through the gaps on its own.

03:32Right? How far can it think ahead without being guided? So that's what we're testing today.

03:36Alright. So the first app that I tried to build is this mobile app for iOS. This is basically like a quiz app with flash cards and study notes.

03:44Currently, it's making around $3,000,000 a month. So I'm gonna try to get Fable five to one shot this app. So here's the prompt that I gave it.

03:51So built a mobile iOS style application called CocoQuiz, a flashcard quiz similar to Quizlet, including the color palette and typography and the layout as well. And I just gave it a bit more context about the features that I wanted.

04:03Also to pre populate the app with some dummy data so we can see if the app is working or not. And I also ask it to build the whole thing in HTML so I can easily preview it on my Chrome browser. And I gave the same exact prompt to Opus as well to make it a fair comparison.

04:18Right? Now if you notice, I didn't give it too much detail here. It's just a very short prompt with just the basic features and a reference app that I wanna copy because the point of this test is to see how far it will go in building out the features and whether it's gonna get it correctly.

04:33So here's what I got. So the one on the left is built with Opus and the one on the right is built by Fable.

04:40So at first glance, I think both of these look quite good. One obvious thing that I noticed is that Opus actually gave me this iPhone frame while this one does not, but I can easily ask it to do that as well. But let me see if these are all clickable.

04:54So these filters are just dummy data, but let me see if I can click on this. Okay.

04:59So this one works. So if I click on flash cards and let me see if this flips. Yep.

05:05So the answer is behind this flash card. So the animation works and if I go to the next one, that works as well. And let's see if I wanna go to quiz.

05:14Okay. So there's a dummy quiz that I can do. Alright.

05:17So I do like the animation and if I go to carts, maybe I go to home and then gonna add a new one. Okay.

05:24Let's do coding, coding. Let me see if it actually works. Maybe do advanced coding.

05:30And then let's hit create. Alright. So the functionality works here, which is nice.

05:35And let's see what we have here. I think both of these, they look equally good. They look very similar in terms of design, but this one they added the icon here because I asked it to name the app Coco Quiz.

05:47So added this icon here. Let's study. There's a fire animation here that is glowing, so I added that.

05:54And if I click on this, yeah, it's very similar to that one and there's a flash card as well which works. I feel like the animation is a bit better here. So it went a bit further in terms of the animation.

06:05So let's click on okay. So it added a feature that this one didn't have which is matching the matching game here.

06:12So if I go back yeah. This one only has flash cards and quiz but Fable was able to create a third feature which is this matching game here. So you can drag and drop all this.

06:22Right? And it has a timer as well. Nice.

06:24So I didn't even ask it to do this but I think this is really really good. Now can Opus create this feature as well? Of course, if you ask it to do it but the fact that it did it without me asking just to think about the features that I might miss, I think that's something impressive.

06:38And I think design wise, both of these are very similar. But if I had to pick one, I'd probably pick this one because the animation is better. But again, all these things you can actually prompt it to add whatever animation you want, but what we're testing today is how far it's able to go with just one prompt.

06:55Alright. So the second app that I built was this web application, basically a booking app like Calendly. I wanted to see if I can just one shot the app as well with one single prompt.

07:04So here's the prompt that I gave it. Build a fully responsive single page web application called calendar with two r's, a scheduling tool similar to a calendar. So I gave it a reference as well, but I asked it to come up with its own design direction and also gave it a couple of features that I think is pretty basic for a calendar booking app like this, like a booking page where users can select a date from calendar and then also a dashboard where the host can manage the bookings and maybe reschedule things like this.

07:33Right? And again, very short prompt and I gave the same exact prompt for Opus as well and this is the result that I got. So on the left here is the one that's built by Opus and on the right is from five.

07:46Now I think that looking at the design here, they both are okay, but the only thing that I don't like about the left one here is the fact that it's using the purple theme here which screens vibe coded app. Right? So I don't like that.

07:58But to be fair, I did not give it any design direction. I didn't specify the color that I wanted. But with Fable, it was able to be a bit more creative and used a different theme which I personally think that it's a lot better than this one.

08:11Now with Opus, there's also a bit of design errors here, so I think the age here is a bit too close to 2.5. But in terms of the functionality, if I click on the reschedule here okay. So that works.

08:23I can pick another time. If I click on two, confirm time. Yep.

08:28So that actually changed as well. And let's say if I toggle Monday off and save it, and then if I go to book a time, this is the page where people can book a time slot with me. Yeah.

08:40So that works, and I can see that I can book a time. Right?

08:45Okay. So I I obviously, I can't book a time yet because it's just dummy data, but functionality wise, it works.

08:53Yep. Forty five minutes, this has changed. Yeah.

08:56To be fair, there's also I think these are too close together but can easily be fixed as well. Okay.

09:03So yep. I think I just like this color better than this one. And then if I go to this one, yeah, I can cancel it.

09:10Can reschedule as well. Yeah. So I think this is good from just one prompt.

09:15I wouldn't say that Fable is better than Opus for this particular app because I'm sure that I can build a much better of this if I give it a more detailed prompt. So I don't think it's worth the double token usage if you wanna build something like this. Now the third thing that I wanted to build was just a simple landing page for a software product like WhisperFlow here.

09:34So if I was building a voice dictation tool like this, I will probably need a website like this to showcase the app. So here is my prompt for Claude. So build a fully responsive single page landing page for voice flow, a voice dictation software that lets users speak and have their words instantly transcribed anywhere on the computer.

09:53And again, I ask it to come up with its own design direction, and I ask it to add some basic actions like the hero section, testimonial section, a pricing sections with two plans and a footer. So very, very basic.

10:06And then I said the page should have smooth scrolling and at least one subtle animation. So that's it and this is what we got. Alright.

10:12So on the left again is the one from Opus and on the right is the one from Fable. Now looking at the left one first, I think it's quite impressive what it's able to build with just like a one paragraph prompt. It was able to build all this animation here and it understood the software or the app that I'm trying to build and created this animation to match the app that I'm building.

10:33Right? So if I scroll down here, it's got the feature section and then the testimonials and also the pricing that I mentioned that I wanted and then a photo section here.

10:42Now the only thing that I don't like about this is obviously the purple color, which is a dead giveaway that it's a vibe coated website. But with Fable, it was able to avoid giving me this color and and added some creativity here, and I think this looks a lot better.

10:56Although, we do have the same kind of similar animation here. But if you notice here, the top is kind of messed up here. Maybe it's the divider here.

11:04Yeah. So if I open it up a bit more, I think it works perfectly. Yeah.

11:08So this one is also the same case. But let's scroll down here and see if it's any different from Opus. So I think this is a little bit more thought out design wise.

11:17I definitely will pick this one over the one that Opus created. And let me scroll a bit more. Okay.

11:22So the testimonial section is a lot nicer as well and then FAQ as well. If I go to the just the top navigation bar here, one, two, three, four, five, we've got five sections and Opus only gave me three.

11:36So we added more sections here which might or might not be a good thing depending on the software that you're building. But it's just nice to see more options here and then I can decide whether I wanna keep it or I wanna remove it. But I think for this one, I would give it to Fable in terms of the design.

11:51But to be honest with you, if I were to design just a landing page like this with no back end functionality, I'm not gonna pick Fable because it's gonna burn through my tokens a lot quicker.

12:01I can give it more design direction to Opus and probably achieve the same thing. Alright. So the next thing that I built here was a game.

12:08Right? It's a it's a simulation game. It's similar to like a SimCity, if you guys are familiar with that.

12:12And I wanted to do this because I wanna stress test the model to see how detailed it can get. Because what I'm trying to build here is basically a city simulation where there's gonna be buildings, gonna be houses, there's gonna be parks, trees, and cars, and people as well.

12:25And I want this to be as real as possible, meaning like I want the traffic lights to work, I want traffic jam, and I wanted to mimic a real city as close as possible. So I gave this prompt to both Fable and Opus, and this is what we got. So on the left here, you have Opus, and on the right is Fable.

12:41So let's take a look at what Opus built first. So at a glance, I think both of them look very similar, but when you zoom into the details here, you can see that all the buildings look really good. Right?

12:51We've got cars here. We've got people walking. The cars are actually boxes here, but with names.

12:57Right? And then and you can also adjust the time here so you can make it faster, you can make it slower, and you can also zoom in and out here. And if you notice here, when it turns to nighttime, it will actually turn on the lights in the building so that is something clever.

13:10Now if you zoom in here, actually the traffic lights are here. So if it's a green light, the cars will actually go and then if it's red, they will stop and create a traffic jam here and people can cross. So that's really good.

13:22But now if you compare this to the one that Fable built, right, so I'm like, I'm gonna make this bigger, you'll see that this is more elaborate. Right? Not only that it built a similar city, but actually built a better one.

13:34So you can see all the buildings here like a financial district in the middle, and you can see houses here. Whereas if you look at the one that was built by Opus, they're basically all the same buildings.

13:45Right? So I would give that to Fable here and if I zoom in and look at the detail here, the cars actually look like cars and they're not just boxes like the one in Opus and people look like people as well.

13:59And you can actually hover on the people here and it will actually say that, okay, this guy's out on a stroll, this guy walking off from a long meeting. So it's very detailed and this is actually fun to watch.

14:12Right? And I also noticed that the traffic lights, they work as well. And when it's a green light, the cars will go and if it's a red light, it will create a traffic jam as well.

14:21Very similar to what Opus built on the left here. But I feel like this is a lot better, a lot more interesting to watch than this one. And the city feels a lot more lively than the one here.

14:30It's definitely a lot busier. And one more detail that I noticed is that you can see the sun actually moving here on top and there's some clouds here while this one actually does not. Right?

14:39So I think that's an interesting detail that it added. So I think from this example alone, we can see how far or or how many extra layers Fable was able to think through when building something like this.

14:52But here's my personal take on when to actually use Fable. Right? Because if you're building something simple like a landing page or a basic website with just some text and buttons, I think Opus 4.8 will give you basically the same result.

15:05There's no real reason to, like, spend twice as much, but when you're building something with a lot of moving parts like maybe a Mac desktop app, which is quite complex to build, or a CRM, I think that Fable is the right choice. I I it thinks a couple of layers deeper and I think that actually matters when things get really complicated.

15:22Now for me personally, I'll still be using Opus for most of my projects. I think the only time when I'll be using Fable is probably when I'm planning something or when I'm mapping out how an app should be structured, that's when I'm gonna use it. And once everything is planned out, I'll probably hand off the build to cheaper models.

15:38Now I think the other way that I might be using Fable for is to help me catch things that I might miss security wise when I'm building something. So did I accidentally expose my API keys or is my database secure?

15:50Is there any security gaps that I should know about? Right? So this is the kind of thing that a model like Fable is really good at flagging and I would use this like sort of like a final check before I launch anything.

16:01So tell me in the comments if you guys have been using this and what you think about it. I'm definitely gonna do more testing and build more apps using Fable, so make sure you're subscribed if not already so you don't miss these videos when they come out. But thanks again for watching, and I will see you in the next week.

The Hook

The bait, then the rug-pull.

A new model drops and everyone runs benchmarks. This one ran four real apps instead — same prompt, two models, side by side — to find out exactly where the extra cost changes the outcome.

Frameworks