YouTube Instagram TikTok How it works FAQ

The fastest way to turn talking videos into clean, searchable text. Free, private and no account required.

About

Built by Rishav Raj with one goal: free transcripts for everyone.

rishav@yttotranscript.com

Transcribe

YouTube to text
YouTube with timestamps
Instagram Reels
TikTok videos

Learn

How it works
FAQ
YouTube transcripts guide

More

Chrome extension
Contact
Privacy policy
Terms of service

© 2026 YTtoTranscript. Made for creators & researchers.

Not affiliated with YouTube, Instagram or TikTok. Respect creators’ copyright.

Claude Opus 5 vs GPT-5.6 Sol: Same App, Same Prompt — Transcript · YTtoTranscript

Home/Transcripts/YouTube

YouTube Transcript

Claude Opus 5 vs GPT-5.6 Sol: Same App, Same Prompt

Code with Beto · 2,214 words · 11 min read · EN-US

Watch on YouTube

Share X LinkedIn WhatsApp

Below is the complete, readable transcript of Claude Opus 5 vs GPT-5.6 Sol: Same App, Same Prompt by Code with Beto on YouTube. Read the full text, copy any part you need, or generate a transcript for any video with our free tool.

Claude Opus 5 just dropped and it is Anthropic's direct answer to OpenAI's GPT-5.6 Sol on Anthropic's own benchmarks it sits at or near the frontier but the price is the headline same input price as sol with cheaper output that sounds great on a chart but i want to know if it can build

something real GPT-5.6 Sol and Fable 5 already built this exact application same plan same starting project same skills and now opus 5 gets the exact same challenge with one prompt so today we can answer the question benchmarks cannot which model should actually build your application let's get into it so i'm going to use the same

prompt i used to build the application with Fable 5 so this is the video in case you want to check it out but most importantly i have a document that you can reference which contains the plan that we're going to use today so let's copy the plan i'm going to go to Claude

i already created habits opus 5 it's a React Native Expo application blank it only includes an App Icon skill since i'm going to require the agent to come up with a name for the application and an app icon so for that it's going to use my skill App Icon you can learn more in the

description in case you want to use the skills and it's going to use as well my MCP server again you can learn more in the description let's make sure that we are selecting opus 5 with the Ultra Code thinking effort this is because in my test with GPT-5.6 i used the maximum thinking effort

available so to make things fair i'm going to use the same i'm going to use the bypass permissions since i want the agent to handle automatically everything without me having to approve everything manually i'm going to paste the prompt and hit enter just like that now while opus 5 is building the application let me

Transcribe another video

Paste any YouTube, Instagram or TikTok link to get a free transcript.

Free · No sign-up · Unlimited

Add to Chrome — transcribe in one click

give you more context about Claude Opus 5 it's basically designed to be used every day and it's now the new default model on Claude Max and the strongest model available in Claude Pro which is the plan that i have the $20 per month now this table is very interesting and it actually got my

attention because i couldn't believe that opus 5 is literally 50% cheaper than Fable 5 and it's better at coding according to this graph and today we're going to actually confirm that but not only Fable 5 which is the most expensive model available at the moment it's also way better than opus 48 so

i already made videos creating apps with Fable 5 i guess you don't have time to watch the other videos let me just tell you that in my test between GPT-5.6 Sol and Fable 5 GPT-5.6 Sol was able to deliver a better developer experience better application and it was able to make the application work

the first time versus Fable 5 no it wasn't able to do it although the end results were very close in that specific test GPT-5.6 Sol won because it's 50% cheaper but today we're comparing it with opus 5 which it's basically the same price it costs $5 in and $25 out per million token versus

GPT-5.6 Sol cost $5 in and $30 out so it's a bit more expensive so according to this table if we compare opus 5 versus GPT-5.6 Sol it's pretty much better at everything except for agentic coding everything else opus 5 is supposed to be better they said that it's better at verifying its work finding root

causes and iterating until something actually succeeds quick update we are 10 minutes in and opus 5 is using the MCP correctly it's gathering information it decided to name the application Tally which is the same name that opus 48 chose when i was doing my test with the same prompt gathering information about the packages

how it's going to implement haptics and it's also researching how to implement the widgets for that since it's pretty new i provided the documentation it's searching the web and so far everything looks great quick update we are 22 minutes in and opus 5 is running the application on my physical device so that's a plus let me

allow this and press reload let's see what we got so far boom the application is working and i believe is compatible with light mode as well this looks really nice now before we start testing let's wait for opus 5 to finish and then we can judge okay guys this is new i haven't seen

this before so claude can now use the iOS simulator and this is new that's pretty cool and it's running it here this is huge because now opus 5 has context on how the application actually looks like for example it's taking screenshots as you can see here so this screen is broken and it's detecting it so it's

saying that the summary block is overlapping the header and now it's rewriting the home screen okay guys so opus 5 finally finished and i gotta say i'm very impressed with all the testing that it did now that you can run the simulator inside claude code you can actually see what it's doing what it's

typing what is screenshotting and validating that it works but you have to keep in mind that it's going to take a while as well for example every time that it took a screenshot it has to view what's in that screenshot and then based on that try to validate and there are a couple of limitations

especially since this is running on a simulator it could not test the widget it couldn't activate it for example the test notification it couldn't unblock the simulator after testing the push notification there but we can see it it couldn't test zoom transitions drag to reorder those gestures are more complex of course and

for that we need to do test manually now before the final test on a physical device let's actually talk numbers because it cost me $73 for this session only and it took an hour most of the time was spent validating that the application actually worked so it also takes time when it builds the application

because it takes like two to three minutes to compile it that said at the end of the day $73 for this specific session is a lot of money especially since when i built the same application with Fable 5 i only spent $50 so you can see here that the total time was an hour and six

minutes and it's roughly the same that it took Fable 5 and GPT-5.6 Sol so on time we are at the same and here we can clearly see that most of the context went into using the iOS Simulator MCP so context window right now it's sitting at 55% of the 1 million context window and

it's also worth noting that i was able to do this thing just with my $20 per month subscription so although the session cost 73 i didn't pay that i am subscribed with the Pro plan and it's not using my usage credits okay time for the real test i have here my physical device and let's see if

it actually works on the first try let's open it we have the onboarding of course we have a dark mode we have a blue as the primary color let's continue okay so it's i guess customizing it because now it's green and now it's orange so let's enable notifications amazing now the title as you

can see there is not compatible with light mode but this is actually an annoying caveat that even experienced engineers need to handle with Expo router but everything else looks great navigation feels great let's load this sample data here we have it so let's go back and take a look the header animation is not working let's

take a look at sorting items ooh that works really nice perfect let's mark this as done we have the animation and it looks great nice background with shadow let's take a look at the history screen i really like it it really seems like the GitHub heat map which is what i was going for as the

inspiration and i specifically asked for a GitHub-style heat graph in the prompt so this looks really nice the header animation is not working so a little detail there send that as notification and here we have it so local notifications are working great so the new habits screen looks really nice and it's more minimalistic

i would say let's say subscribe subscribe and add it okay so here's the item now let's take a look at the actions if i swipe left i can edit and i bring it here or i can delete it all right at this point the application actually works but if i were to actually launch this

app building it would only be half of the job the next problem is figuring out how to get people to discover it for this video i partnered with Virlo Virlo lets you search any niche and track creators across short form content you can see real-time data AI reports and creator outliers from TikTok YouTube

Instagram and the Meta Ad Library without having to manually scrape every platform and because Virlo connects directly to Claude Code through the MCP i asked opus to research content formats already working for habit and productivity applications Virlo analyzed hundreds of videos and slideshows and the result generally changed how i would launch this application

dedicated habit tracker demos were not as strong as opportunity Virlo found 61 videos around iphone home screen aesthetics and widgets compared to only 15 dedicated habit tracker demos so instead of marketing my app as another habit tracker the stronger angle is a widget that makes your home screen look good which also happens to track your

habits opus then turned that research into three original launch concepts a home screen carousel a silent eight second demo and a simple builder story asking viewers for feedback and that is what makes this workflow useful i'm not asking AI to guess what might work i'm giving it real social data and using it to create an original

launch plan if you want to connect Virlo to Claude, Codex, or your own product check out the link in the description now before i launch anything there's one test left does the widget actually work now the real test to me is check if the widget actually works so let's take a look oh it is not

displaying anything so that's kind of weird just so that you get a sense of how this looks on different applications so the one at the bottom was created by Fable 5 the one in the middle was created by GPT-5.6 Sol and the medium size doesn't look great but the small one and large one looks great

that's how it looks and seems that opus 5 wasn't able to test this of course i'm pretty sure that it would have been able to fix it if it was able to take a screenshot of it but it wasn't while developing it so it couldn't test it so i'm not going to create a follow-up

prompt because it's going to be very expensive i would actually recommend that you create a new chat after the first long session but overall i'm pretty happy with the results so can Claude Opus 5 beat GPT-5.6 Sol focusing only on the product the final result i would say this is a tie for this

specific application what worries me a lot is the cost because $73 for this session it's a lot of money now on the other hand this is included in the Pro plan so i didn't pay for that it was able to test it inside Claude Code application by taking screenshots and iterating although that also consumed

extra context so if you're trying to use less tokens maybe just don't use the simulator because it's going to use a lot of context window and it would also take more time to deliver the first version so right now of course i'm also using Ultra Code which is going to take more time to reason

seems to me that GPT-5.6 Sol is more efficient and the final results were pretty much the same but let me know which model you're going to be using if you want to see GPT-5.6 Sol go through the exact test that video is right here watch it next and compare what it builds before you

choose a model for your own application and thanks to all the channel members for your support it really helps me keep creating content like this and paying the AI bill if you want to get early access to future videos and exclusive content consider becoming a channel member see you in the next one