19:19Dubibubi · Review
I Made Claude Opus 5 and Fable 5 Build the Same App
A blind, three-build test of Anthropic's newest coding models — and the first time this reviewer picked a model other than Fable as the winner.
July 25thThe argument in one line.
Claude Opus 5 matches or beats Fable 5 on real one-shot coding tasks while costing half as much per token, splitting a two-test head-to-head 1-1 on raw output quality but winning decisively on price.
Read if. Skip if.
READ IF YOU ARE…
- You use an AI coding agent (Claude Code, Cursor, or similar) and are deciding which model to default to for build tasks.
- You care about cost per token as much as raw output quality when picking a coding model.
- You're curious what a single, unedited one-shot prompt produces across competing frontier models.
- You're already on Claude Max or Pro and weighing when Opus 5 beats sticking with Opus 4.8.
SKIP IF…
- You want a rigorous, multi-prompt statistical benchmark rather than one creator's two anecdotal tests.
- You don't use AI coding tools at all.
The full version, fast.
A creator sends the identical prompt to Fable 5, Opus 4.8, and the newly released Claude Opus 5: first, build a data-driven website about the post-AI economy; second, build an endless flight simulator over a city. Opus 5's economy site is dramatically more polished than both competitors, with a working 3D globe and accurate animated stats. On the flight simulator, Fable 5's city is denser and more detailed, with sound, so it wins that round. The tests end 1-1 on quality, but Opus 5 is priced at $5 per million input tokens and $25 per million output tokens — the same as its predecessor Opus 4.8, and half of what Fable 5 costs — so the creator's conclusion is to default to Opus 5 for daily agentic coding work.
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 →Where the time goes.
00:00 – 00:52
01 · Opus 5 launches at half Fable 5's price
Cold open recaps Anthropic's Opus 5 release: $5 per million input tokens and $25 per million output tokens, the same price as Opus 4.8 but half of Fable 5. The host calls out that Opus 5 is the new state of the art per Anthropic's own claims, with one exception: it stays behind on cybersecurity tasks.
00:52 – 01:43
02 · The case for switching to Opus 5 daily
The host argues there's little reason to keep defaulting to Fable 5 when Opus 5 offers comparable intelligence for half the cost, framing this as the emerging consensus in the agentic coding space, and asks viewers what further model comparisons they'd want to see.
01:43 – 02:46
03 · Test 1: the same prompt, three models, Fable 5 goes first
The rules are set: the identical prompt — build a web app about the state of the economy after AI — goes to Fable 5, Opus 4.8, and Opus 5. Fable 5's one-shot build (finished fastest, in about 7 minutes) has a polished 3D background and stat sections, but the host flags a visibly misaligned 3D element as unexpected from a frontier model.
02:46 – 03:00
04 · Opus 4.8's version: solid but not close
Opus 4.8's build of the same prompt has similar animated sections and 3D touches, but the host says it's 'nowhere close' to how good Fable 5's build looks.
03:00 – 04:52
05 · Opus 5's build: 'a different league'
Opus 5's one-shot economy site draws the strongest reaction of the video: accurate statistics, a professional-looking layout and fonts, an interactive 3D globe visualizing where frontier AI research and compute cluster geographically (Bay Area, Beijing, Bengaluru, London), and animated bar charts with glow effects the host calls better than both other models.
04:52 – 06:14
06 · Test 2: the flight simulator — Fable 5 wins this one
Second head-to-head: an endless flight simulator flying over a city. Fable 5's build has a denser, more colorful skyline with lights, tall structures, and working sound. Opus 5's version, while functional with plenty of buildings, is described as visually more basic and less detailed — evening the overall score to 1-1.
06:14 – 07:52
07 · Verdict: split on quality, decided on price
The host recaps both tests, lands on the two models being roughly equal in output quality across different task types, then repeats Opus 5's pricing and Anthropic's benchmark comparison as the deciding factor — half the price of Fable 5 for comparable performance — before closing with a request for viewer input on future model comparisons.
Lines worth screenshotting.
- Claude Opus 5 launched priced identically to Opus 4.8 — $5 per million input tokens, $25 per million output tokens — which is half the per-token cost of Fable 5.
- Anthropic's own release notes place Opus 5 at the new state of the art on its benchmarks, with one stated exception: it remains behind on cybersecurity tasks.
- Given the same single prompt, three models produced visibly different one-shot builds: Fable 5's had a 3D element that was visibly misaligned, Opus 4.8's was serviceable but plain, and Opus 5's was described as 'a different league.'
- Opus 5's economy-site build included a working 3D globe showing frontier AI research clustering in specific regions — Bay Area, Beijing, Bengaluru, London — pulled from a single prompt with no follow-up iteration.
- On a second, harder test — an endless-flight city simulator — the result flipped: Fable 5 produced the more detailed, more colorful city with working sound, and Opus 5's version was comparatively basic.
- Across two different one-shot build categories, the two models split 1-1 on output quality, suggesting neither is a strict upgrade over the other on raw capability alone.
- The deciding factor the creator lands on isn't which model 'wins' a given test — it's that Opus 5 delivers comparable results for half of Fable 5's price.
- Fable 5's one-shot economy-site build was the fastest of the three to complete, at roughly seven minutes.
How to actually compare two AI coding models
MODEL SELECTION
A single one-shot prompt run through competing models can flip winners between task types, so the real decision usually comes down to price for comparable output, not a single test's winner.
- The same prompt can produce very different results across models and across task types — a model that wins on one build type (data-driven website) can lose on another (physics-heavy game simulator).
- Don't judge a coding model from a single one-shot test; run it against more than one kind of project before deciding it's better or worse than an alternative.
- When two models produce roughly comparable output quality, price per token becomes the practical tiebreaker for which one to default to.
- Read a vendor's own benchmark claims for the caveats, not just the headline: Anthropic's own release notes flagged Opus 5 as behind on cybersecurity tasks specifically, even while claiming state-of-the-art elsewhere.
- A model that finishes a build fastest isn't necessarily the one producing the highest-quality result — speed and polish are separate axes worth checking independently.
Terms worth knowing.
- One-shot build
- Giving an AI coding model a single prompt and evaluating whatever it produces without follow-up corrections or iteration.
- Agentic coding
- Using an AI model inside a tool that can write, run, and modify code autonomously across multiple steps, rather than just answering a single question.
- Vibe coding
- Building software by describing the desired outcome in natural language and letting an AI model generate the implementation, with little to no hand-written code.
Things they pointed at.
Lines you could clip.
00:00
“Anthropic just released Opus five... and the results will shock you.”
03:00
“This is absolutely a different league.”
06:14
“Fable five and Opus five are quite equal in a lot of different ways... but the key difference here is the pricing.”
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.
Anthropic just released Opus five, and in this video, we've vibe coded applications with Opus five and then built the same applications with Fable five and Opus four eight. And we compared those applications, and the results will shock you.
Make no mistake, the release of Opus five is the most important model release of this entire month. Now in terms of pricing, Claude Opus five is the same price as its predecessor Claude Opus four eight, which is a very big deal. So it's $5 per million input tokens and $25 per million output tokens.
Alright. So that's that for pricing. It is half the price for Fable, which is absolutely incredible.
And this is Anthropix release about Opus five. I'm not gonna go over the benchmark because I don't believe in benchmarks, but what it's saying here is that Opus five is the new state of the art, though it remains behind Methos five on cybersecurity tasks. So it is their best model, but for cybersecurity task, it is still behind Methos five, at least according to Anthropic.
So you're saying Opus five is designed to be used every day, and this is something I agree with. Why would you use Fable five for the majority of your tasks when you can get the same level of intelligence and performance for half the cost. This is the consensus currently what I'm hearing in the agentic engineering space that you should probably start switching over to OPUS five because you're getting the highest intelligence for a fraction of the cost of what Fable five is.
So this is a decision that we're gonna see being made over the course of the next few weeks. We're gonna see people using this in their daily workflows, and then we'll be able to understand more about this.
So if you guys are interested in videos on that, let me know in the comment section down below. If you guys wanna see more tests, maybe comparing OPUS five to different models, maybe like Kimi k three, just let me know in the comment section down below and I'll release that in just a few hours. Let me know.
That's it for this video. Thank you guys for watching and peace. Alright.
So the rules for this test are simple. We've given the same prompt to Fable five, Opus four eight, and Opus five. Build us a web application talking about the state of the economy after AI.
So this first one right here was built by Fable five. You can see that there's some really nice three d in the background. I can scroll down here.
This is pretty nice. This is a decent website. Fable five is known for this.
This is a really nice animation we just saw here, so it's saying that more jobs have been created than displaced by AI. Um, so this is all very nice. You can see that there's some three d here, but this is a little messed up, which is not expected from Fable five given that it is at the frontier of intelligence.
But right here, you can see that the three d is misaligned, is something that we might have to iterate on. So that is not great from Fable five.
Everything else seems pretty decent. Alright. So this was Fable five, one shot, and this was the fastest.
It took seven minutes to build. Then we have Opus 4.8. This is what Opus 4.8 built.
It has some nice three d as well, but it's definitely nowhere close to Fable in terms of how good it is. So you can scroll down here. You can see that it is a pretty decent website, some nice animations here as well.
Nice three d, a shift, not a cliff. So it's talking about the same topic.
Alright. And then what really surprises me here is what Opus five built and you will instantly see the difference. So I head right over here and this is what Opus five has built and this is absolutely a different league.
This is actually even better than Fable five. This is much better than what Fable five built. Right?
So I can scroll down here and you can see that this is very beautiful. This looks like a professional website and wow, this is absolutely beautiful. So the statistics are accurate.
You can see I can move this three d right over here, which is really nice. It's really comfortable compared to not having that in other, uh, in Opus four eight or Fable five. And you can see the text here.
The fonts actually look very professional. I can scroll down. The fastest growing occupations.
The three d here is absolutely incredible. This is so much better than Fable five and Opus four eight. Opus five is completely winning here.
This is actually really, really surprising. So you can scroll down here and wow.
This is absolutely incredible. Diffusion of AI across the firm. You know, the three d here is incredible.
The glow is incredible. The design is definitely goes to OPUS five.
The design definitely goes to OPUS five. This is absolutely incredible. So it is not happening everywhere at once.
Research capital and compute cluster tightly. A relatively small number of metropolitan regions host the majority of frontier model development. That is absolutely true.
San Francisco Bay Area is probably the number one spot right now for that. So this is absolutely incredible. Um, we can actually maneuver here so we can scroll around.
We can see that Beijing is pretty important here. Um, Bengaluru is pretty important. So we have all of these.
London is important. Right? So we have a really nice three d here.
What I'm trying to say here is that OPUS five definitely wins here. Right? This is probably The United States.
Yep. So that's a $109,000,000,000, The United States, um, whereas China and The United Kingdom are much far behind.
And so this is absolutely incredible. You can see that the three d here is just completely on a different level. This is nowhere near Opus four a.
This is just a completely different league. And even Fable five is not competing here. This was the exact same prompt, and yet Opus five absolutely crushes both those models.
Alright. So for the second test, we've built a flight simulator with both Fable five and OPUS five. And the one that you're seeing on my screen right now was built by Fable five.
So while OPUS five may be the underdog, Fable five definitely seems to be better here. And I'm gonna show you what Opus five built in just a minute, but you can see here that Fable has built a very nice city and an endless flyover of the city. So we have nice skyscrapers, we have buildings, we have Ferris wheels, we have lights all over the place.
This is a very nice flyover. I can go down. I can go up here and I can view this this way.
I can view this this way. And it has sound as well. I'm not gonna toggle that for now.
And then what, um, Opus five has built is this, which you can see is not as good as what Fable five is built, at least in my opinion. I feel like the city is not as detailed. Um, there's not as much color.
It's a much more basic city. But there are a lot of buildings, um, if that's something you wanna consider.
But overall, in terms of design, I would say that this is much better. Fable five definitely wins here. So what this goes to say is that Fable five and OPUS five are quite equal in a lot of different ways.
So the first test, we saw that OPUS five clearly won. And in the second test, I'm pretty sure that Fable five clearly wins. And this is the consensus in the agent decoding space that both these models are performing equally well.
Um, sometimes Fable five gets better, sometimes OPUS five performs better, but the key difference here is the pricing. OPUS five is half the cost of Fable five, and that is a very, very important thing. So you get the same performance as Fable five, but for half the price, and that is a game changer.
Now in terms of pricing, Claude Opus five is the same price as its predecessor Claude Opus four eight, which is a very big deal. So it's $5 per million input tokens and $25 per million output tokens.
Alright. So that's that for pricing. It is half the price for Fable, which is absolutely incredible.
And this is Anthropix release about Opus five. I'm not gonna go over the benchmark because I don't believe in benchmarks, but what it's saying here is that Opus five is the new state of the art, though it remains behind Methos five on cybersecurity tasks.
So it is their best model, but for cybersecurity task, it is still behind Methos five, at least according to Anthropic. So you're saying Opus five is designed to be used every day, and this is something I agree with. Why would you use Fable five for the majority of your tasks when you can get the same level of intelligence and performance for half the cost.
This is the consensus currently what I'm hearing in the agentic engineering space that you should probably start switching over to OPUS five because you're getting the highest intelligence for a fraction of the cost of what Fable five is. So this is a decision that we're gonna see being made over the course of the next few weeks.
We're gonna see people using this in their daily workflows, and then we'll be able to understand more about this. So if you guys are interested in videos on that, let me know in the comment section down below.
You If guys wanna see more tests, maybe comparing Opus five to different models, maybe like Kimmy k three, just let me know in the comment section down below, and I'll release that in just a few hours. Let me know.
That's it for this video. Thank you guys for watching, and peace.
The Hook
The bait, then the rug-pull.
Anthropic just released Opus five, and in this video, we've vibe coded applications with Opus five and then built the same applications with Fable five and Opus four eight.
How they asked for the click.
VERBAL ASK
07:22next-video
“If you guys wanna see more tests, maybe comparing Opus five to different models... just let me know in the comment section down below and I'll release that in just a few hours.”
Soft, low-pressure ask for comments to steer a follow-up video rather than a product or subscribe pitch.
MENTIONED ON CAMERA
Visual structure at a glance.
