ChatGPT Astra ist raus! 😮

Chat 6, specifically Astra, is partially out, and it's supposed to be better than Fable 5.1. What's the deal with it? It was announced today, Friday , or rather last night, from Thursday to Friday. Over the last few weeks, it has been shown and leaked multiple times, and it's always the case now when big models come out that a few companies get early access. And so, a lot of things have already been circulating in this, let's say, hardcore AI community. The benchmarks came out yesterday, and we'll take a look at them together; specifically, here on Artificial Analysis, which is the main benchmark everyone is looking at right now, and it’s very disappointing here. You can see GPT 6 has 61 and Fable 5 has 66. So, in terms of intelligence, Claude is still significantly better, even now with two models. I have to say, I was really disappointed by that. It was speculated beforehand that it would be over 70, so actually a huge jump, and now it's just one or two points better at most compared to 5.6, the previous model. So I have to say, that's pretty disappointing, and in terms of speed as well as cost per task, it's actually gotten worse. But one thing, and that's why I can't wait to use GPT6, has become really insanely good, and that's the topic of power use and computer use . That is also the reason why I now use Codex more than Claude Code. Simply put , it's about AI agents being able to use your computer, and that’s where we see the biggest leaps with GPT6. The benchmarks say, and several users who already have access now confirm, that it is roughly—when you tell it to adjust something on your PC, meaning when it virtually takes over your keyboard and mouse—three times as fast and also significantly more accurate. And that is the actual big game changer for GPT6, that it is supposed to be much, much better when it takes over your computer or your browser. So, it's a bit like an intern trained to take over your computer; it's not much more intelligent, which has always been my personal opinion. The models are already so smart, I don't need that; instead, I need more speed and for them to operate my computer better, so they can take more tasks off my hands, and that's what it should be. As soon as it's available to me, I will make another detailed video about it. I think even on YouTube, to show how it will change my everyday work life. Yeah.