Did 5.6 Sol Just Close The Fable Gap?

**Brian** (0:00)
Hey, what's going on, everybody? Welcome to The Daily AI Show. Today is July 10th, 2026 It's 10 a.m. on the East Coast, 7 a.m. Pacific. And right now, I'm joined by Andy. I think we'll have Beth here. And I think I just maybe saw a message that said Karl Beall to pop in for a little bit as well. So it should be a Good Friday show. Not Good Friday, but a Good Friday.
Not that time of year yet. But yeah, I think there's, you know, Andy, I think one of the things people probably wanna know the most about is what really ended up being a fairly decent announcement yesterday from ChatGPT. And that's back to back. You know, we knew we were expecting information as far as 5.6 Sol, Terra, Luna. The day before we had gotten, and we talked about it yesterday in the show, the GPT Live 1, which was, we talked about it a lot. It's, my experience has been very good. Beth was saying she had sort of a intermittent experience as far as the interrupting part of it. But overall, I think it's a really powerful model. And in fact, I just saw them do a live event where, for work, which we're about to talk about, where they did live translation. I'm going to assume there was no magic there and that it was the actual live translation. But the gentleman was talking to somebody who is only speaking Japanese, and they were using the ChatGPT Live mode, voice mode to do all that. And I'm going to assume it was live because it did mess up at one point. And I think the guy panicked. The ChatGPT, the OpenAI guy panicked. And he said, no, ChatGPT, can you translate that? He kept saying Japanese. And so what he was trying to say was English. He was like, no, can you translate to English? And I don't know if you saw this or whatever. And ChatGPT is like, sure. I can't speak Ohio Gazanmas. I don't know how to speak Japanese, right? So like whatever it was. And he's like, no, no, no, Japanese.
And he's like, yep, that's what I'm doing. And so he stopped and started the app again. I just thought it was more comical because I think the audience is like, you're saying the wrong word. And he's like, whatever, I'll just stop and start it. But anyway, it was very impressive. So all that to say, that's where we were coming into yesterday, but then they did a live announcement of GPT Work.
And there's a lot of different pieces to this Andy, right? That are worth talking about because GPT Work has made some big changes to Codex. I think some people are upset about that from the early conversations I've seen, that what you do on mobile versus desktop versus browser are different and have different pieces and parts to them.
And overall, it's probably what we suspected yesterday when we were speculating about what work would be, is that it's definitely a move away from this chat, text in, text out. It's much more about workflows and these types of things, and ChatGPT doing work on your behalf because it can connect to many different types of systems from your Google drives to so on and so forth. So that's just sort of a starting point. But Andy, I have more obviously on what work is and we should get into the details. But did you have an opportunity to even see the announcement yesterday? Or have you done that recently?

**Andy** (3:17)
I've read the blurbs from the various newsletters about work, but I'm very interested. I hadn't looked at that myself and I haven't opened the Codex app this morning to see what it looks like in there. So I'm very interested in your demonstration. But let's talk just quickly for a minute about 5.6 Max, which is what was made available to all the testers and everyone else yesterday. And let me share my screen and I'll just show you what Artificial Analysis says about it. Give me a second here.
This is the Artificial Analysis Coding Agent Index. And you'll see here that the big news is that it now, state of the art, is 5.6 Sol, their leader. But matching Claude code Fable 5 is Terra.

**Brian** (4:12)
Sorry, I should have put it back on. We're good. It was back on the screen. Sorry about that, Andy. Go ahead.

**Andy** (4:18)
No sweat. No sweat. So the first point I wanted to make is that Codex GPT-56 using Terra, the middle range model, matches Fable 5 on that Artificial Analysis, multi-benchmark aggregated index. So you can basically say that OpenAI achieved what their needed gain was against the massive attention that came to Anthropic in their side-by-side race to the finish here for the IPOs later this year or earlier this year.