If you've been on X lately, you've probably heard of Claude code and you may have heard of Codex from OpenAI, but you're probably unsure about which one you should use, which one is more expensive, and what are the benefits and drawbacks of using each model? In this video, I'm going to answer all those questions, and you might be surprised at my ultimate recommendation. So first of all, what is cloud code?
What is Codex? If you already know, you can skip ahead, but I'm gonna give a quick summary. Claude code and Codex are basically coding agents.
So they are AI agents that are specially trained to write code. Claude code is from Anthropic. If you've used Claude before the Claude app that you see right here is where .
The, the Claude Chat is cowork, which is a new AI coworking feature that is meant for business tasks. I've created some videos on that and then the coding tab, you can also use Claude Code in the terminal. But it works really nice in the CLO apps.
So that's what I'm gonna show you. Okay, codex, on the other hand from OpenAI is its own app. You don't have chat GBT and Codex in the same app.
They're, they're different apps. But it works in a very similar way. In both of them, you're going to open a folder on your local computer.
It's going to work in your local computer, and then you will be able to write code locally. But of course you can always sync with GitHub. So you can commit your code to GitHub, to a GitHub repository.
So all your code is in GitHub. In the cloud but also on your local computer here, unlike tools like Lovable and Bolt, which have integrated backends with Claude Code and Codex, you have to integrate that on your own. So you can integrate something like Supabase to do that.
That's what I typically do when I'm building an app for myself. And if you are curious about how to do that, you can always just ask the agent, how do I integrate with Supabase? And it will explain it to you, but, but you have to set up your own backend and user authentication.
So these are the apps. There's the Claude app, and this is Claude Code. And then you have Codex.
This is the Codex app. So you have the apps, but then you also have the models that are running inside the app. So right now inside of Codex, the latest model is GPT 5.
4, and then you can select the reasoning between low and extra high. Now I typically like GBT 5. 4 on extra high.
Then you can also do a slash command to go into some different modes. And, and do skills as well. So for example, you can turn fast mode on and off.
Fast mode basically just increases the speed of your response, but you have to pay two x the tokens. You can also go to plan mode, for example which is a really nice mode that both code X and clo code half. Over in the cloud code app, we have Opus 4.
6. That's the, the newest, the most powerful, newest model sonnet. And Haiku.
So if you're coding, I generally recommend using Opus 4. 6. The way to think about it is I think it's worth using the most powerful model 'cause you're gonna end up spending more time and more tokens to go back and fix code that didn't work, as opposed to just having code that works the first time.
I generally find Opus 4. 6 is better for using for, for design and user interface and GBT 5. 4 extra high I found to be really good for more complex tasks that need to be done methodically.
Now both of these are really good models and can complete both of those types of tasks, but if you're using both that's my personal finding. But again, some people might have different findings. The models are very dependent on the context of what it's working on the prompting that you're giving it and all of that.
The next thing to talk about is pricing. So, codex pricing is currently, you can try it for free with Chad, GBD, free or go and pay $20 or $200 to get more usage. I'm on the $20 a month plan on Codex and I have never run into any limits right now.
They are giving a lot of leeway with limits 'cause I think they want people to try it out and switch over from Claude Code. That could change in the future, but right now I definitely recommend getting the $20 a month plan and you're gonna get a lot of usage out of that. On Claude on the $20 a month plan.
This is annual, but it's $20 a month. You're gonna get a, a good amount of usage but you're probably gonna need to upgrade to at least a hundred dollars, the a hundred dollars plan. To do anything substantial.
I kind of think I would kind of compare it as okay. Right now the. A hundred dollars on Anthropics Max plan is similar to the amount of coding I get out of the $20 a month.
Open AI plan. You can also upgrade to $200 on the max plan to get even more usage. But, you know, I recommend starting lower and then if you hit your limits, you can always upgrade.
So there's no reason to upgrade until you're hitting your limits. But generally I'm recommending. If you're new to this to try out Codex first, GBT 5.
4, I think you'll get more value for your money with the $20 a month plan. And I think that the models are really good. Now I'm gonna show you a quick example.
I put the exact same prompt into both of these models. Now this is just one simple example. So of course it could be, you're gonna get different results from, for all sorts of different outcomes.
But here's the snake game that Claude created and something nice about the Claude Code app actually is that it, it has a preview tab here so I can preview my app inside. I can't do that with Codex right now. But yeah, it works.
It's, you know, pretty basic, but it does work. And then here I did the exact same thing. I'm running a local server here and I'm doing that with both, actually a local server.
I just asked each of the models how to do that and it'll do that for you and run the commands. And yeah, here we are with with the Codex version of this and let me know in the comments which one you think is better and if you have any opinions on, which of these interfaces you like more. The AI world is changing so quickly.
So be sure to subscribe to our YouTube channel to stay up to date with future updates on everything that's happening. Even these models themselves. This is the newest model, GBT 5.
4. This is the newest opus 4. 6.
I wouldn't be surprised to see Anthropic release and new opus model soon, and I'll make a video on that and compare that to GBT 5. 4. I also created an AI newsletter that I read every morning to stay up to date on everything I need to know as an AI developer.
So click the link in the description to subscribe for free and stay up to date. Thanks for watching. We'll see you in the next video.
