GPT-5.6 is HERE!

**Julian Goldie** (0:00)
GPT 5.6 is here, my friends, but it's not for us. So, what's happened here? This was a new report by Axios that came out today. Basically, the Trump administration asks OpenAI to limit the next model release. And this is pretty big news. So what this means essentially is that it is limited to government-approved partners before any wider release due to security concerns. And if you haven't seen the news already, basically, we're looking at a new reality here where basically, you wouldn't get access to the new stuff that comes out. So if you're wondering, for example, this is a great article by Silicon Valley. And this is not confirmed officially, but it is confirmed from some reports, for example, with the information.
And basically, what's happened is the government in the US has asked OpenAI to stagger the release of new models due to the Fable 5 situation. So if you weren't aware of that already, basically what happened with Fable 5 is it got taken down after like three days. And it was an amazing model, but it was great while it lasted, and then it kind of disappeared and nobody can access it again. It was absolutely crushed and all the benchmarks had created some amazing stuff. So what we're looking at here, as you can see in this article, is that we are looking at a situation where GPT 5.6 is only available to a small group of partners. Now, this is pretty wild. I've never seen this before. Essentially, what you're looking at is a reality where the US government can say, okay, you can give this model, but only to specific people, and then it won't be released to everyone until it's fully tested. And so it kind of like runs through a filter first where the US government would approve it, then they approve anyone who can use it, and then it doesn't roll out to everyone until at least a couple of weeks after, and even then it may not get released at all. And I think this has all come from the Fable 5 situation where Claude's model got taken down. So something that's interesting, this is on Thursday apparently, CEO Sam Altman told staff that the government will be approving access to GPT 5.6 customer by customer.
There's no open application for this, you can't just say, hey, I want access to this, here's my name, here's my details, etc. Really what we're looking at is a situation where basically, due to the powerful cybersecurity capabilities of these models, especially with MIFAS and Fable 5 as well, this would be a staggered release, and it's not going to be out for everyone. Now, if you look at the notes, and again, this is not confirmed by Sam Altman, this is an article from Axios and also from The Information. So they've said the government, apparently in this internal memo from Sam Altman, the government would be approving customer by customer access during this preview period with a wider roller hoped for. And the key word is hoped for. So there's nothing confirmed that would actually come out later. We may never even get access to a level of technology that was the same level as Fable 5 or anything like that. And so, you know, you've got GPT 5.6 can't release, it's ready to ship. The government check in, there's actually two federal offices that would vet it. It would be approved one by one, customer by customer. And then wider release would be a couple of weeks later, if not way longer, right? And this is really the first time a US model has got a queue. You know, we're used to AI launches being instant. GPT 5.5 comes out, boom, the whole world's talking about it, the whole world's got access to it within hours. Whereas if you're looking at this situation, really this breaks the rhythm. And for the first time, a major US model is shipping behind a government managed line. So apparently what's happening is the really this is probably only going to get access to like cyber security firms, right? So kind of like Project Glasswing, which is where Clawed and Anthropic basically said, okay, not everyone's going to access this model. We're going to keep it behind closed doors. It's too powerful to release.
It's too much of a frontier level AI.
We can't release this to everyone. Now, I'll be honest with you as well. Like when I was testing out Fable 5, like sometimes it would actually make mistakes. So even if this model gets released, I don't think it's going to be changing the world anytime soon. It may be a big step up from GPT 5.5, but it's not going to be released to everyone. And I think this is probably a pattern that we can expect from pretty much every single frontier model that comes out from the US particularly. Now, there's two angles to this. Number one is like, in a way, this is good, because you know, it protects people, it protects the internet from any dodgy stuff, and everything is carefully vetted before it gets out of control. In a way, it's pretty bad for the US, because if you imagine, for example, like GLM 5.2, which is pretty much Claude Opus 4.8 level, if not better, if you've got these Chinese models like that, being open source, being released, coming out pretty much every single month. So I wouldn't be surprised if GLM 5.3 comes out next month, and it's even better than 5.6.