Fable 5.1 Just Dropped. It Looks Unreal.

0:00 Well, it's here. We have Claude Fable 5.1 and Mythos 5.1. They are the world's most advanced models for coding and knowledge work. I mean, right now, Fable 5 is definitely the most capable model

0:09 that I've ever played with. And 5.1 is apparently much, much better and a little bit cheaper. Now, when I say cheaper, I don't mean the input and output cost. It's the exact same $10 per

0:19 million input and $50 per million output, but it seems to be more efficient, which ultimately means it eats less tokens and costs less. And obviously, it literally just came out,

0:26 but it is available right now. You can see that I've got some running right here in my desktop app, Fable 5.1. And I'm just comparing Fable 5.1 and Fable 5. So, trust me, I'm going to have a

0:35 video coming out later after I've gotten time to play with them for a few hours and see how they feel and see how they cost. But you can see we're starting to get some outputs back. So, maybe at the

0:43 end of this video, I'll compare a few things. But anyways, right here is the actual blog, and it looks different. It looks a little bit different than the way their blogs typically look. We can

0:50 change the color up here. We can change the the theme, but you'll notice in the bottom right that it says that this was made by Fable 5.1. So already it just feels a little bit different and I'm

0:59 really interested to see how Fable 5.1 does with design. And you can see right here at the top they already say Fable 5.1 will cost an estimated 25% less than Fable 5 for typical workflows wherever

1:09 usage is built by token. This is because we're reducing our pricing on cache reads where the model reads input that we've already processed and it stores it. And for highly agentic work, the

1:17 savings will often be much larger up to approximately 50% would be insane. They also have a new EFS system which you can read about on the separate page which basically lets enterprise customers opt

1:27 in to have you know no data retention being stored. It's a complete privacy policy which is obviously great for these enterprises that are looking to adopt Claude. And they've also improved

1:36 the safeguards to reduce false positives. This is pretty interesting too because literally just a few hours ago I dropped a video and like 12 hours ago they dropped um their blog on what

1:46 they were doing with an older claude model. Basically training it to be evil. training it to hack. And so if you guys haven't seen that video, I'll tag that right up here. Definitely check it out.

1:54 It was an interesting one. But after they have done all that research, now they come around and say, "Oh, by the way, also here's Fable 5.1 and we have improved our safeguards to reduce false

2:02 positives and making this stuff more safe. Now look at this. A new Frontier performance. Now these graphs are really interesting. Now obviously take these metrics and take these benchmarks with a

2:11 grain of salt. You have to get in there, see how it works. But look how much better this is showing that Fable 5.1 is compared to Fable 5. Now, not only better on the effort, like Fable 5.1 on

2:22 low is better than the strongest Fable 5 on max or high or X high and it's cheaper. Right over here, we can see $1110. And over here, we see 44 bucks or we see 34 bucks. This is incredible. And

2:35 if this really holds true, then it's going to get really fun to get in here and play with Fable 5.1. Same thing on agentic terminal coding. You can see that it's very similar pattern as well.

2:44 Multidisciplinary reasoning and agentic coding as well. Even this chart, like it's nothing crazy, but just the way that it's designed definitely feels different than the previous blog posts

2:54 of model releases where they typically used like Fable 5 to design Fable 5's release blog and Opus 5 to design Opus 5's release blog. And you know, it wouldn't be an enthropic blog model drop

3:06 if there wasn't the best GBT model in here as well, which obviously in these benchmarks, it looks like it's crushing it out of the water. But this is a massive jump over Fable 5. Like that is

3:15 really interesting. Now we can see here Claude Mythos 5.1 is identical to Fable 5.1 but it offers more permissive safeguards for vetted individuals and organizations whose work is affected by

3:25 the cyber security and life sciences restrictions outlined above. So it's only available Mythos 5.1 is only available to two trusted access programs. The cyber verification program

3:34 and the life sciences verification program. So if you fit in there, go ahead and apply. You might as well get your hands dirty with Mythos 5.1. And once again, this stuff is available

3:43 today. You can see I've got it running right in here in my cloud desktop app. You could also be running this of course in the terminal or I think it said via API as well, including Amazon Web

3:53 Services, Google Cloud, Microsoft Azure. Developers can get started with cloud-f API. So, real quick before I end the video, let's take a look at these results. We can see here that I've got

4:03 Fable had built me a 3D bear. Here it is. You can see we've got a bear on a bike. We can move it around. I can click space to pause. I can click space to play. It is nothing crazy. Obviously, my

4:14 prompt here was literally just, "Hey, build me a rotating 3D cartoon bear riding a bike." And you can also see that Fable 5.1 finished it up. It was the exact same prompt and it gave me a

4:23 live artifact. It also named the bear Barnaby. But let me open this up real quick. You can see that we have the bear. I mean, this obviously looks a bit better. You can see the other one was a

4:32 little bit more colorful, but the actual bear itself and even kind of the physics looks a little bit more realistic. Um, like I said though, this is not any sort of like definitive test or benchmark,

4:42 but the bearer definitely looks better in this one. There's like the shadows. You can see the shadows look way more realistic. The physics feel more real. Um, but that's a super open-ended

4:50 prompt. Now, real quick, what I want to do is do a slash cost. And we'll see how much this one spent. This one costed us um about 6 minutes 34 seconds and $453. If I go back into the Fable 5 version,

5:04 we'll do the same thing. Slash usage. Shoot that off. This one was $541 and it was a little bit faster, but already on a very simple example, Fable 5.1 was roughly a dollar cheaper for this

5:16 example. And I've got these other experiments running that are going to take a little longer. I'm going to design a bunch of websites. I'm going to try their things and I'll make a video

5:22 later basically telling you guys my findings. But anyways, that is going to do it for today. So, I hope you guys enjoyed the video or you learned something new. Please give it a like.

5:28 I'll see you guys in the next one. Thanks everyone.