Why Modern Creator?

Duncan Rogoff | Learn Claude Code · YouTube

Claude Opus 5 vs Fable 5: A Live Head-to-Head Test

A Claude Code YouTuber runs Opus 5 and Fable 5 through the same three-part marketing prompt to see if Opus 5's half-price tag actually pays off.

Posted

July 24th

1 months ago

Duration

09:59

Format

Review

hype

Views

20.5K

255 likes

Part of the collectionThe Fable 5 PlaybookAll 45 Fable 5 breakdowns, synthesized into one page.

Read the playbook

Part of the collectionThe Claude Opus 5 PlaybookEvery Opus 5 breakdown, synthesized into one page.

Read the playbook

Big Idea

The argument in one line.

Claude Opus 5 launched at roughly half the per-token price of Fable 5, but in a live three-task marketing test the two models ended up costing and taking almost the same amount, so the real differentiator was output quality, not the rate card.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You're choosing between Claude Opus 5 and a competing frontier model and want a real cost/quality comparison instead of a benchmark screenshot.
You run an agency or freelance shop building landing pages, email sequences, or ad scripts for local-business clients and want to see what AI-generated output actually looks like.
You're curious whether Anthropic's benchmark claims for Opus 5 translate into visibly better real-world creative work.

SKIP IF…

You want a rigorous, controlled benchmark methodology — this is one anecdotal run of one prompt on two models, not a study.
You only care about raw coding/agentic benchmarks with no marketing-copy use case.

TL;DR

The full version, fast.

Claude Opus 5 launched at $5 per million tokens versus Fable 5's $10, while Anthropic's own benchmarks show it matching or beating Fable on agentic coding, knowledge work, and novel problem-solving. The host ran both models on an identical three-part prompt — build a roofing-company landing page, write a five-email cold outbound sequence, and generate a 15-second video ad — in two side-by-side terminals. Opus finished in 7:43 for $5.16; Fable finished in 6:20 for $5.93, so cost and time came out close despite the 2x sticker-price gap. On the landing page and email copy the host called it a toss-up, but preferred Opus 5's video ad for its stronger story arc.

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 – 00:14

01 · Opus 5 Just Dropped

Cold open: the host claims Opus 5 already beat Fable 5, minutes after launch.

00:14 – 00:46

02 · Opus vs Fable Pricing

Anthropic's pricing page shows Opus 5 at $5/M tokens versus Fable 5 at $10/M tokens.

00:46 – 01:04

03 · Setting Up Both Models

Two terminals opened side by side; Opus 5 set as default on the left, Fable 5 on the right.

01:04 – 02:03

04 · The Head-to-Head Test Prompt

Host reads the identical three-part prompt: build a Summit Roofing landing page, a 5-email cold outbound sequence, and a 15-second cinematic video ad.

02:03 – 02:39

05 · Opus 5 Benchmarks

Anthropic benchmark table compares Opus 5, Fable 5, Opus 4.8, and GPT-5.1 across agentic coding, knowledge work, novel problem-solving, and agentic search.

02:39 – 03:01

06 · Cost vs Effort Levels

Chart shows Fable 5's higher-effort tiers catching up to Opus 5 in score but at significantly greater cost.

03:01 – 03:41

07 · Novel Problem Solving Wins

Host highlights Opus 5's novel problem-solving benchmark lead and a clip of it building a physics-based wind tunnel.

03:41 – 04:21

08 · Why Opus Checks Itself

Host reads customer testimonials crediting Opus 5's edge to repeatedly checking and validating its own work before finishing.

04:21 – 04:55

09 · Comparing Cost and Runtime

Reveal: Fable 5 cost $5.93 in 6:20; Opus 5 cost $5.16 in 7:43 — close despite the 2x price-per-token gap.

04:55 – 05:31

10 · Opus 5 Landing Page

Walkthrough of Opus 5's Summit Roofing landing page: hero form, proof points, 4-step process, reviews.

05:31 – 06:25

11 · Fable 5 Landing Page

Walkthrough of Fable 5's version: brighter illustration-led hero, services-first layout, contact form at the bottom.

06:25 – 08:14

12 · Cold Email Sequence Compared

Both models wrote the same subject line for email one; Opus 5's body ran longer and more technical, Fable 5's was shorter and more conversational.

08:14 – 08:53

13 · Fable 5 Video Ad

Fable 5's 15-second ad plays: push-in shots of a roof, a couple of AI mispronunciations, ends on a CTA card.

08:53 – 09:40

14 · Opus 5 Video Ad

Opus 5's ad plays: a full storyboard-driven mini-story of a roof failing and a crew arriving, which the host preferred.

09:40 – 09:59

15 · Which Model Should You Use

Sign-off: host asks viewers which model and outputs they preferred, plugs a related prompting video.

Atomic Insights

Lines worth screenshotting.

Claude Opus 5 launched priced at $5 per million tokens, half of Fable 5's $10, marketed as comparable intelligence at half the cost.
On CursorBench 3.2 at max effort, Opus 5 scored within 0.5% of Fable 5's peak while costing half as much per task, per Anthropic's own charts.
Opus 5 more than doubles Opus 4.8's performance on Frontier-Bench v0.1 software-engineering tasks at a lower cost per task.
Despite a 2x per-token price advantage, Opus 5 took longer to finish the same real task than Fable 5 — 7 minutes 43 seconds versus 6 minutes 20 seconds.
In a live build of a roofing landing page, email sequence, and video ad, Opus 5 cost $5.16 and Fable 5 cost $5.93 — nearly identical despite the 2x per-token price gap.
Anthropic's cited customer testimonials credit Opus 5's edge to checking and validating its own work repeatedly before declaring a task done.
The one area Fable 5 reportedly still leads is advanced cybersecurity tasks like finding vulnerabilities in open-source code.
Given the identical prompt, Opus 5 and Fable 5 independently wrote the exact same cold-email subject line: 'Quick question about your roof.'
Fable 5's landing page leaned on reassurance ('the Arizona sun is brutal on your roof — we're not') while Opus 5's leaned on urgency ('your roof won't survive another Phoenix monsoon').
Opus 5 generated a full storyboard with image prompts before rendering its video ad, which the host suspects explains its longer runtime and slightly higher cost on that task.

Takeaway

A live model bakeoff beats a benchmark screenshot

AI MODEL BAKEOFF

Testing two AI models on the same real deliverable, not just reading their benchmark charts, is the only way to know if a cheaper model actually costs you less in practice.

02Opus vs Fable Pricing

Opus 5 launched priced at $5 per million tokens, half of Fable 5's $10 per million tokens, positioned as the same intelligence at half the cost.

03Setting Up Both Models

The fairest way to compare two AI models is running the identical prompt in parallel, in two terminals, at the same time.

04The Head-to-Head Test Prompt

The test prompt bundled three distinct deliverables into one request — a landing page, a cold email sequence, and a video ad script — to stress-test follow-through across formats.
Real deliverable prompts (client-ready copy and pages) reveal quality gaps that abstract coding benchmarks don't.

05Opus 5 Benchmarks

Anthropic's own benchmark charts show Opus 5 leading on agentic coding, knowledge work, novel problem-solving, and agentic search against Fable 5, Opus 4.8, and GPT-5.1.

06Cost vs Effort Levels

Raising a model's 'effort' setting trades more inference spend for higher benchmark scores — Fable 5's high-effort tier catches up to Opus 5 but at significantly greater cost.

07Novel Problem Solving Wins

Opus 5's standout benchmark category was novel problem-solving — tackling new, unscripted challenges rather than well-covered task types.

08Why Opus Checks Itself

The recurring theme in Anthropic's cited customer testimonials wasn't raw task completion — it was Opus 5 checking and validating its own work before declaring the task done.
A model with fewer cases of misaligned behavior than competitors is a reliability edge that doesn't show up in a single-run demo.

09Comparing Cost and Runtime

Despite a 2x per-token price gap, Opus 5 and Fable 5 ended up costing almost the same on the real task — $5.16 versus $5.93 — because Opus 5 used more tokens to finish it.
Opus 5 also took longer end-to-end (7:43 vs 6:20), so a cheaper rate card doesn't automatically mean a cheaper or faster job.

10Opus 5 Landing Page vs Fable 5 Landing Page

Opus 5's headline led with urgency ('your roof won't survive another Phoenix monsoon'); Fable 5's led with reassurance ('the Arizona sun is brutal on your roof — we're not').
Opus 5 put the contact form directly in the hero section, removing a click between interest and lead capture; Fable 5 saved the form for the bottom of the page.
Given an identical brief, the two models made different design calls — worth generating both and picking, not settling for the first draft.

12Cold Email Sequence Compared

Both models wrote the identical subject line for email one ('Quick question about your roof'), suggesting some cold-outbound patterns converge regardless of model.
Opus 5's email ran longer and more technical (explaining underlayment lifespan); Fable 5's was shorter, plainer, and read better to the reviewer.

13Fable 5 Video Ad vs Opus 5 Video Ad

Fable 5's ad was competently shot but had noticeable AI mispronunciations ('shingers' for shingles, a garbled 'monsoon').
Opus 5 generated a full image-prompt storyboard before rendering its video, which likely explains its longer runtime and slightly higher cost on that task.
Opus 5's ad won out because it told a complete story (calm day, roof fails, crew arrives) rather than a sequence of disconnected, well-shot scenes.

Glossary

Terms worth knowing.

Opus 5: Anthropic's Claude model referenced in this video as newly launched, priced at $5 per million tokens.
Fable 5: A competing frontier AI model referenced in this video, priced around $10 per million tokens.
Agentic coding: A benchmark category measuring how well a model can autonomously write, run, and fix code across multiple steps without step-by-step human guidance.
Effort setting: A model configuration (e.g. low, high, max) that trades more inference cost for higher output quality on a given task.
Misaligned behavior: Cases where a model does something other than what the user actually wanted, used here as a reliability comparison metric between models.
Higgsfield AI / Seedance: Third-party AI video generation tools the host had each model write prompts for, used to produce the video ad clip.

Resources

Things they pointed at.

00:14linkAnthropic Opus 5 pricing / announcement page ↗

01:31toolHiggsfield AI

01:31toolSeedance

00:00productLearn Claude Code (Skool community, $9) ↗

Quotables

Lines you could clip.

00:00

“Opus five just dropped, and I think it killed Fable, and I don't even think that's an exaggeration.”

bold cold-open claim→ TikTok hook↗ Tweet quote

02:50

“So really it just kind of seems like we should all just be using Opus for everything at this point.”

strong mid-video verdict line→ IG reel cold open↗ Tweet quote

04:40

“Fable costs $5.93, and Opus costs $5.16.”

the numbers punchline that undercuts the 'half the price' pitch→ newsletter pull-quote↗ Tweet quote

09:55

“I don't know why everyone's walking backwards, but that's kind of beside the point.”

self-aware humor as a closer→ 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.

story

Quad Opus five just dropped, and I think it killed Fable, and I don't even think that's an exaggeration. So in this video, we're just going to talk about Anthropic's latest model, Opus five, which literally just dropped, like, ten minutes ago. I'm actually going to compare the two models head to head, Opus versus Fable on a couple of different builds so we can really test them out and see which performs better.

So as of today, Opus five is available to everyone, and it basically has the same level of intelligence as Fable five and at half the price. And if you come over just to the pricing right here, we can see that Fable five is $10 for a million tokens, and OPUS is only $5 for a million tokens. So we're literally doing the same thing, if not better, at 50% of the cost.

So let's just test this out head to head, OPUS five versus Fable five on a couple of different examples. So I just have a couple of terminals opened up here, and I'm just gonna type in model. And you can see we already have access to OPUS five right here.

So I'm just gonna set this to the default, and I'm gonna come over to this one. And on the right over here, we're gonna make this our Fable model. Because these models are so good at these complex tasks or multistep tasks, this is just kind of like a prompt I wrote to test it on a couple of things.

So let me just read this to you. You are a marketing agency hired by Summit Roofing, a roofing company in Phoenix, Arizona. Do three things.

One, build a complete landing page for them as a single HTML file, modern, mobile responsive, conversion focused, realistic copy. Include the hero section with headline and CTA, services section with four services, testimonial section with three reviews, contact form, and footer.

Two, write a five email cold outbound sequence for targeting homeowners who need roof replacement. Include subject lines, output as a CSV with columns, email number, subject line, body. And then this last one I just changed to say write a fifteen second cinematic ad with time stamps and prompts for each shot and generate the video using Higgs Field AI and Cdance.

And then 16 by nine, four eighty p format just to keep it cheap, and then this is just kind of like a little description of what the format should look like. So it says do all three, output each clearly labeled with a separator line, and I'm just gonna copy this, come back over into Opus five on the left here, and paste that in.

And on the right hand side here in Fable, I'm gonna do the same. We're gonna see how long they take. We're gonna see how much they cost, and we're going to check out the overall results and quality at the end.

And so we can actually see that even just comparing the benchmarks, like, Opus five, like, greatly outperforms Fable. And it's pretty crazy because the Fable hype was, like, real for the last, I don't know, three, four weeks or so.

And so I'm, like, pretty impressed at how much better, like, Opus seems to be already. It clearly wins at things like terminal coding, knowledge work, novel problem solving, agentic search, and pretty much everything else up and down this list.

And we can see even just by looking at this chart here that this yellow line is Fable, and so we can see like the higher end or higher level of efforts of Fable will outperform Opus, but at like significantly greater cost. And so really it just kind of seems like we should all just be using Opus for everything at this point.

One thing Opus absolutely dominates at is novel problem solving, which is basically just this idea that you give it some sort of new or innovative idea and Opus will figure out how to achieve it. We can just watch this clip really quick where apparently Opus five built a working wind tunnel with like real world physics.

Like, that's pretty incredible, honestly. In terms of performance, Quad OPUS five provides greatly improved performance for the same cost as its predecessor, OPUS 4.8. OPUS five excels on valuable software engineering tasks, and what's pretty crazy is that it surpassed all other models and more than doubles OPUS four point eight's performance at a lower cost per task.

And then at max effort, OPUS five performs within point 5% of Fable five's peak score, but at half the cost per task. And it also achieves greater performance than all other models on high, extra high, and max effort.

I went through and I read a couple of these little testimonials or reviews, and really the common theme is this idea of Opus checking its own work. The thing that stood out to all of these major companies was not only Opus five's ability to execute on these complex tasks, but it was the way that it checked and validated its own work before completion.

So OPUS is actually going to do the work and then check itself repeatedly until it is positive it has the result that you're after. Compared to all of the other models, it has significantly less cases of misaligned behavior or basically doing things that you don't want it to do.

And it really seems to outperform all of the other models including Fable five on pretty much everything except for these like advanced cyber security tasks like finding all of these vulnerabilities. Okay. So both Opus five and Fable five finished their tasks.

We can see they each built a landing page. They built a cold email sequence and a cinematic ad. So we're going to review all three of these to see which came out the best.

In terms of cost and time, I actually found this to be interesting is that they kind of cost similarly. Like, Fable costs $5.93, and Opus costs $5.16.

Opus took a little bit longer to run, seven minutes and forty three seconds compared to six minutes and twenty seconds. So I'm not actually sure why Opus seemed to be doing so much more work to complete this task. We'll never know.

But I always like to give it to you straight, so that's what it was in this instance. So this is the first example of our landing page for Summit Roofing Company. This was the one done by Opus five, and we can see this looks pretty good.

I'm pretty happy with this. I love that there is a form on the right side here where somebody can just come put in their contact details and book their inspection. We can see we have some proof points and some stats.

We can actually see the services that are provided down here. Like, I think this looks pretty nice. We actually see a four step flow of how this whole thing works, so nobody is confused on the process.

We even included some five star reviews across the board. And one last CTA at the bottom here. This looks pretty good.

I'm generally pretty happy with this site. I really like the look of it and the type of information that it provides to a potential customer. Now this is the version that Fable made.

Totally different look. It has this more, like, sunny bright vibe for Arizona and roofing, obviously. So we can just go back and compare the headline.

So the Arizona sun is brutal on your roof. We're not. And if we look back at what Opus wrote, says it your roof won't survive another Phoenix monsoon.

So this is really digging into fear, and then this is sort of like playing up the capabilities of the roofing company. One thing Fable did was insert this cute little house graphic, which I kinda like. You can see what we do.

So this is the the services immediately here. No proof points initially like the other one had. We have them down here, which looks nice.

Couple of reviews, and one contact form at the bottom. And so, really, it kind of just comes down to taste and personal preference.

I like both of these sites. I kinda do like having the contact form right at the top there if you're just kind of ready to book. This is, again, the one by Opus and the one by Fable.

So let me know in the comments which one of these sites you prefer. So the next thing to compare is these cold email sequences. I'm obviously not going to read through all of these, but I just kinda wanted to check them out to see.

So email number one, this is done by Opus. Subject line is quick question about your roof on and then the street name. So it's already personalized.

And here we can see that this is the actual email message. So hi, first name. I'm Ray with Summit Roofing here in Phoenix.

We're replacing a roof over on a nearby street last week, and I noticed a few homes in your neighborhood are hitting the age where the underlayment starts to go. Most Valley homes built in the late nineties and early two thousands are on their original underlayment. Tile lasts fifty years.

The felt paper under it lasts about twenty in this heat. That's the part that actually keeps water out, and you can't see it all from the ground. Not pitching you anything.

I'm just offering a free thirty minute inspection. We get up there, photograph every slope, and tell you how many years you've got left, etcetera, etcetera.

So, though, I think this is generally pretty good. This is probably too much context and too wordy in here and too technical for most people. But, again, this is what Opus created.

I like that it has this little sign off down here, and I'm generally pretty happy with this. Although, personally, I may make a few adjustments. So this is what Fable five created.

So subject line is quick question about your roof on and then the street name. So are these actually exactly the same, which is pretty cool? Yeah.

These are exactly the same, so I think that's interesting. And it does look like the email itself is significantly shorter. So hi, first name.

I'm Mike with Summit Roofing here in Phoenix. We've been doing work near street nearby and noticed most roofs in your area went on fifteen to twenty years ago. The Phoenix sun is hard on a roof.

Around year fifteen, shingles start to dry out and crack even when things look fine from the street. We do a free fifteen minute inspection with a photo report so you can see exactly what's up there. No pressure if your roof is fine.

I'll tell you it's fine. Want me to swing by this week? Reply here a call.

I actually like this email significantly better. It's simple. It's straight to the point.

Not a lot of fluff or extra detail. So this is the email that Fable wrote, again, this is the email sequence that Opus wrote. So you can judge for yourself which of these you like better.

And the last thing to compare is our two video ads. So let's watch the Fable five ad first.

The Arizona sun is brutal on your roof. Cracked tiles, curling shingles, it only gets worse.

Summit Roofing replaces it right in days. Built for Montesquieu season, guaranteed for life.

Summit Roofing, book your free inspection today. So I actually kinda like that. A couple of mispronunciations, Shingers and monsoon or something for monsoon.

But generally speaking, like, this is, like, pretty nice. Like, I kinda like the push in. Like, the detail is really nice, and I get it.

Right? So pretty good. Let's check out the one from Opus.

One thing I did notice about the Opus one is that it actually generated this storyboard first, and so maybe that's one of the reasons why it took extra credits because it actually spent time, like, generating some images and image prompts before actually generating the video.

So this is the video generated by Opus, and, honestly, I like this one so much more. The main reason I like this more is that it really, like, tells a full story. Right?

Like, you're on this beautiful sunny day, and then all of a sudden, the roof starts to leak. And then this part here where these badass dudes from Summit Roofing show up in this truck cracks me up, and I think it's great and dramatic and fun.

At the end here, I don't know why everyone's walking backwards, but that's kind of beside the point. So, anyway, Opus five is out today. Let me know in the comments which model you prefer and what you're going to build with it first.

If you wanna learn how to make the best possible prompts for Cloud Code, just check out this video right here. I'll see you over there.

The Hook

The bait, then the rug-pull.

A YouTuber pits Anthropic's freshly launched Claude Opus 5 against Fable 5 in a live, identical three-task test — build a landing page, write a cold email sequence, generate a video ad — to see whether Opus 5's half-price tag actually translates into a cheaper, better result.

Frameworks

Named ideas worth stealing.

01:04list

Three-Task AI Model Bakeoff

Build a landing page
Write a cold email sequence
Generate a video ad script/video

A single prompt asking one AI model to complete a landing page, a 5-email cold outbound sequence, and a 15-second video ad for a fictional local business, used to compare Opus 5 and Fable 5 head-to-head on real deliverables instead of abstract benchmarks.

Steal forquickly bakeoff two LLMs on real marketing-agency deliverables before committing to one for client work

CTA Breakdown

How they asked for the click.

VERBAL ASK

09:52next-video

“If you wanna learn how to make the best possible prompts for Cloud Code, just check out this video right here.”

low-pressure end-card plug to a related video; the $9 Skool community link appears only in the description, not spoken on camera.

MENTIONED ON CAMERA

00:14linkAnthropic Opus 5 pricing / announcement page ↗

00:00productLearn Claude Code (Skool community, $9) ↗

FROM THE DESCRIPTION

PRIMARY CTAWhere the creator wants you to go next.

👉 ↗

Storyboard

Visual structure at a glance.

open

hookopen00:00

benchmark chart

valuebenchmark chart02:47

Opus 5 landing page

valueOpus 5 landing page05:20

Fable 5 ad CTA card

valueFable 5 ad CTA card08:38

sign-off

ctasign-off09:53

Frame Gallery

Visual moments.

open

Frame at 00:07 from Claude Opus 5 vs Fable 5: A Live Head-to-Head Test

Frame at 00:11 from Claude Opus 5 vs Fable 5: A Live Head-to-Head Test

Frame at 00:19 from Claude Opus 5 vs Fable 5: A Live Head-to-Head Test

Frame at 00:26 from Claude Opus 5 vs Fable 5: A Live Head-to-Head Test

Frame at 00:35 from Claude Opus 5 vs Fable 5: A Live Head-to-Head Test

One-click upgrade to your Google

Get more breakdowns in your search results

Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.

Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.

Watch next

More from this channel + related breakdowns.

10:58

Duncan Rogoff | Learn Claude Code · Tutorial

Claude Fable 5 Just Changed Viral Videos Forever

A screen-recorded walkthrough of building a custom Claude Code skill that watches any viral video, breaks it into timestamped beats, and recreates it with Higgsfield's Seedance 2.0.

July 6th

07:16

Duncan Rogoff | Learn Claude Code · Tutorial

Claude Design Is Insane: Pro-Level Websites Instantly

A walkthrough of the /design command now built into Claude's desktop app, from a one-prompt homepage to a reusable design system to multiple creative directions at once.

September 1st

09:27

Duncan Rogoff | Learn Claude Code · Tutorial

Claude Fable 5.1 + Claude Design: Instagram Carousels

A three-level walkthrough of prompting, saving a design system, and layering in AI-rendered images to turn one topic into a branded Instagram carousel.

September 1st

17:45

Duncan Rogoff | Learn Claude Code · Tutorial

I've made $450k+ with AI (here's what I'd do if I was starting from 0)

A former art director for Apple, PlayStation, and Nissan walks through the exact Claude Desktop workflow he'd use to go from zero to a paid $27 product in a single sitting.

August 4th

08:27

Duncan Rogoff | Learn Claude Code · Talking Head + Screen Recording

The New Way Top Anthropic Engineers Prompt Claude Code

Andrej Karpathy says the best way to brief an AI coding agent is a messy ten-minute voice ramble — so a creator tests it live by building a naturopath website from a single, uninformed viewer request.

August 2nd

Video of the Day09:40

Duncan Rogoff | Learn Claude Code · Tutorial

Anthropic Engineers Just Fixed Claude Code, and Nobody's Talking About It

An Anthropic engineer quietly revealed that Claude Code deleted over 80% of its own system prompt — and shipped a built-in /doctor skill that does the same cleanup on your CLAUDE.md, skills, and settings.

July 28th