Well, Fable 5 is back, but there's a catch. Actually, there are several big catches that are going to upset a lot of people. I'm going to go over what those are, and we'll also look at what you can and cannot do with the new Fable 5, and all the restrictions that have been placed on it.
I've been doing some quick testing around it and have some helpful results for all of you. So, first, let's start with some good news. Fable 5 is back.
If you open up Claude, you should see this now. Here in the desktop app, you may see this big message or on the model picker, you'll see Fable 5 in there. But unfortunately, that is all the good news, because even though we have access to Fable 5, it comes with a lot of restrictions.
So first, Fable 5 is only included for up to 50% of your weekly usage limits, which means as soon as your weekly usage limit hits 50%, then you will not be able to use Fable 5. You will have to go back to using Opus or Sonnet. And it's only through July 7th, which is less than a week away.
After that, you will be able to use Fable 5, but it's through usage credits, which cost extra on top of your subscription. So, this is a lot less than we were getting before, because when Fable 5 first came out, we were able to use all of our limits for Fable 5 for a couple of weeks, and then after that, it was moving over to usage usage credits. So, we are getting a lot less than we were getting before, which is really unfortunate.
So, we're not going to have a like a huge chance to try this out and really see its capabilities before it moves over to additional cost with the usage credits. Second, in order to release this model publicly, Anthropic had to increase the safeguards for Fable 5. And they specifically did that using something called classifiers.
These are automated AI systems that, as you are chatting with the model, if it identifies that you're trying to or even potentially trying to use it for harmful cybersecurity tasks, it will block the model from responding. This existed when Fable 5 first came out, but now these safeguards have been increased. So, if we look at this classifier diagram that they've drawn up for us, you can see that for the normal safeguards, they consider a lot of the tasks that you send the model to be benign.
So, those would be allowed. Everything to the right of this dotted line will be blocked. So, there are the things that are clearly harmful, like if you're trying to build complex exploit chains.
There are things that are ambiguous, and then there's this safety margin. They haven't changed what is considered harmful or ambiguous, but that safety margin they've increased by quite a bit, which means that more of your requests and tasks are going to be blocked if you try to use Fable 5 for them. Anthropic has specifically said that this is going to happen more often for routine coding and debugging tasks.
So, if you're a coder, you're using Fable 5 for coding, this is bad news for you. This means that more of the tasks that you're trying to do are going to be blocked. Big picture, this is bad.
Going forward, every state-of-the-art model is going to be scrutinized and gatekept by the government. That precedent has now been set. We already don't have access to the best model, Mythos, and even the watered-down versions are going to be limited for us.
Overall, I think that is very bad. But, let's look at real-world usage. How does this translate?
What is actually getting blocked and what isn't? Well, I've been using the new version of Fable, admittedly only for about 30 minutes now, so we definitely have to do more testing around this, but I started out by giving it a lot of the common tasks that I normally use Claude CoWork for, and those are knowledge work tasks, and it didn't hit a block with any of them. I'm also not a coder, but I am vibe coding some apps as a side project.
I gave Fable 5 access to those files, told it to take a look at them, and tell me where things can be improved, and it successfully did that. No issues there. The only time I was able to trigger a block was when I showed it a Chrome extension I'm working on, and literally asked it, "What are some of the vulnerabilities in this Chrome extension?
" You can see right here, it did answer my question, but it automatically switched from Fable 5 to Opus 4. 8. And you can click here, you'll see why that is.
It says, "Fable 5's safeguards flag this message. The safeguards are intentionally broad right now, and may flag safe and routine coding, cybersecurity, or biology work. These measures let us bring you mythos-level capabilities sooner, and we're working to refine them.
" But besides that, I have not been able to trigger this block in any of my other tasks, and certainly not with my knowledge work. So, that's some good news. The other good news is we have some version of Fable now.
That was what Anthropic had to do to get it back into our hands. Regardless of why it was blocked in the first place, Anthropic could not release it again unless they changed something. Lots of people were speculating that Anthropic might require people to upload IDs to use Fable, and I think we all should be glad that they found a way to avoid that.
So, given these new restrictions, particularly around the usage limits and how short of a time we get access to Fable 5, how should we be using it, if at all? Well, in the limited time that I've spent with Fable 5, I've noticed that it is an incredible model, but it is overkill for most tasks. So, what I'm planning on doing is using Opus for a lot of my knowledge work type tasks.
Then for the most complex ones, I can switch over to Fable 5. And if I limit how much I use it, I think that will be very sustainable. But, just to end on a positive note, if you are using this for knowledge work and you can afford to both pay a subscription and the additional cost of Fable 5 after July 7th, then the good news is that both Opus and Sonnet are great models and you're going to be more than fine with either of them.
So, some good news for Fable 5, some bad news and bad news overall for the entire AI industry. But, guys, let me know in the comment section down below who you think is to blame for all of this drama over the past month. I'd love to hear from you guys and your take on all of this.
Then, like I said before, I'm always staying on top of the latest AI news, keeping you guys updated. So, if you appreciate this video, give it a thumbs up, subscribe to the channel so you don't miss out on future updates, and then I'll see you in the next video. Bye for now.
