Mythos 5 + GPT 5.6 are HERE but…

**Julian Goldie** (0:00)
We have two huge announcements today. Number one about Gpt 5.6, and then number two about Claude Mythos and Fable 5 I'm going to cover them both and keep you updated today. So let's get straight into it. First of all, we have a brand new announcement from Gpt 5.6, OpenAI, which is that Sol, Terra and Luna are released. But biggest issue is they're released in preview. That means me and you, my friend, me and you, we cannot access these models, and it might be a while before we actually get access to them. There's no confirmation or release date, which I'll come on to in a second. But essentially, we could be looking at a minimum a few weeks. They said they're literally hoping for this to come out in a few weeks, but they have no control over this because it's down to the Trump administration and what they think is best at this point. So you can see here for now at the request of the US government, we're starting with a limited preview among a small group of trusted partners in Codex and the API. So what's happening here is they have released it but to like 100 people. That's it. Only 100 people have access to this and it's not 100 people really, it's 100 organizations that are trusted as partners of the US government, I believe. So if we're looking at this, what has happened here, what's been announced? We've got three different models, Sol, Terra, Luna. Now Sol is a flagship model for ambitious agent work. You can see that it costs $5 per input.
Terra is a balanced model for efficient everyday work. That is $2.5, which is 50 percent of the price of Sol. Then we have Luna, and that is at $1 per input. Now at this point, you might be going, what is even the point of this? What is the point of this? We are talking about models that have no release date. We're talking about previews that are new flagship models. We're talking about US government gating AI intelligence. Some people will be watching this and going, well, what's the point of even looking at this? I would look at it and just look at the benchmarks, but at the same time, I would also be looking at this from the perspective of, okay, what's the solution here? Because if frontier models are being gated, what are some ways to get better level intelligence and AI? I'll come on to that in a minute after I've talked about Claude Fable 5, which is the next big update. You can see here inside the benchmarks on Terminal Bench 2.1, Claude Mythos 5 is behind Gpt 5.6, Sol, Ultra and Gpt 5.6 Sol. So we have Claude Mythos 5 over here and we have Claude Fable 5, which is even further down the rankings, which brings me to Fable 5 and Mythos 5 So if you wonder what's going on with them, what is the crack with these guys? Basically, Fable 5 has struck a deal with the Trump administration, according to the CNBC, with Anthropic, which grants a company permission to release its Mythos 5 model.
However, to a group of 100 companies. So again, it's being gated again, and nobody has access. And again, we don't have access to these models, and we are kind of wasting our time trying to find out what's going on and where we're up to. So I'm gonna talk about the best solution to that. And there's basically, I think, three to four very good alternatives to this that will help you a lot. And the first one that I'm gonna talk about is Hermes Agents. Hermes have released something called Mixture of Agents. And if you're wondering what Mixture of Agents is, this is their answer to fixing the issues behind for a gated frontier technology. So what do I mean by that? Well, basically what we're looking at here is one prompt with Hermes Agent, for example, and that will run through a list of models that you can set up personally. I'll show you how to do that in a second. So that, for example, you could have one aggregator, which means you get better answers than if you're using one model alone. So, for example, if you're asking, right, how do I achieve Fable 5 level intelligence without Fable 5? Well, you could use a mixture of agents. You could use Opus 4.8, Gpt 5.5, any other model as reference models inside Hermes. Now, if you're wondering how to set this up, what you can do is make sure you've updated your Hermes Agent. So, for example, we've already updated it inside the agent operating system. Just go to your manage section if you've already got that set up. If not, get it inside the AI Powerful Boarding, link in the comments description. And what you can do from here is you go to models. Now, there are three options. So you have your main model, you have auxiliary tasks, and then you've got a mixture of agents. So you can configure this and you can use a combination of agents together. The whole point behind this is that when you have multiple different models working together, you tend to get better answers than if you have one working alone. So a team of geniuses will probably be one genius alone. So this is option one, which is mixture of agents with Hermes Agents, and it's a new model setting. You can also go inside your terminal here, and if you type in Hermes model, like so, you can scroll up to mixture of agents and then just set it up from there. So that's the option. Next up, there's also something called Fusion, and you'll see a trend across all of this, which is I don't care what model comes out.