ChatGPT Ultrafast: Is It Worth It?

Julian Goldie (0:00)

ChatGPT Ultra Fast just dropped. And basically this means you can use Astra 6, GPT Astra 6, eight times faster than the standard and four times faster than Astra Fast. And I've already upgraded. So we've got the ChatGPT Pro 500 plan, which is the new plan you need to use this. And once you've done that, you can go to chatgpt.com. Then we've got it inside work. And if we go over to our model, we just want to make sure that we have Astra selected. And then if you hover over this, it's actually, it's not obvious. So if you click on this, then you'll see that you've got three options. You've got fast, and then you've got ultra fast. And ultra fast is the fastest of the fastest for using Astra 6 fast. Then from here, what you can do is you can plug in your prompt, and let's just test it out. So we'll take a prompt we used from the Claude Sonnet 5.5 Design Playbook here. We're gonna plug that into ChatGPT like so, using the ultra fast mode.

And now you can see that is running. Bear in mind that as you talk about, I would say 20 minutes to build with Sonnet 5.5. So let's see how quick it is. We can just have it running in the background over there, whilst we do something else. And we can also compare the outputs versus something like Sonnet 5.5. So we're gonna take this prompt, like go back into ChatGPT, and we'll try this too.

And you might also be saying like, is this worth it? Here's a way that I would look at it. Is reverse engineer how much time is gonna save you and whether it's worth it for that amount, right? So based on your time, based on how much you value that, you can easily figure out, okay, is this worth it or not, right? It's eight times faster. Bear in mind that if it's running faster, you are going to use up a lot more tokens. You can also use this on the API. So whilst these are running in the background, we can check out the API system here.

Also, I think something you could easily accidentally do is just leave it on ultrafast all the time. So just be careful that if you start a new chat and you don't need ultrafast, you just want to take it so you don't use loads of tokens up. Also, you might be wondering, is there any quality difference here? So there doesn't seem to be any quality difference, just much faster actually building. We can see the time is here. So three minutes and two minutes over here. It's still building out as you can see. So I just want to be 100 percent honest with you. But it did take about 20 to 30 minutes to build this stuff out. So we'll wait for that to get done. Let's check out on something that we would probably actually use day to day, which is building our website. And what I'm also going to do is test on editing. So I'm going to give it something to edit over here. Also, one thing to note is that it's available in Codex, ChatGPT Work, the API, but it's also coming to GPT 6.1 Sol soon. Bear in mind, I've tested out GPT 6.1 Sol. Wasn't that impressed so far. So I don't think you're missing out on anything if that's not available on Ultrafast. GPT 6 Astro is probably a much better model from my own tests.

In reality.

All right, so we have the first version back, as you can see right here. So this is the 3D planet world. As you can see, you might be wondering how long did that take? So that worked for 5 minutes, 39 seconds. Pretty fast to build something like this.

Honestly, when I look at the outputs, I'm like, this is nowhere near as good as something like Opus 5.5 even using the exact same prompt. But it works and we can move around like you can see right here. I just genuinely think Opus 5.5 is a much better model.

When you compare the outputs here. But yeah, pretty cool. Let's have a look at the next one. So this is a 3D website, or it should be, but it's pretty basic, right? If we have a look through here, again, using Ultrafast. And I don't know if that's just because Astra 6 isn't that good of a model, or whether the quality actually drops because of Ultrafast on my own tests. Apparently, it shouldn't. But so far, I wouldn't use this. I wouldn't use this website, right?