its latest model, the Claude Fable 5, which was a protected version of the Claude Mythos model.
I tested the model day and night since its release, and within 72 hours the model was available.
Unfortunately, the model is not currently available, as it has been suspended by order of the US government. However, since I got to try the model more, I want to give my thoughts on the model, what I was able to do with the model, its limitations, and so on. Although the model is not available at the moment, I believe and hope that it will be available again soon. I also think that in a few months we will see other models with similar skills.
Of course, there have been plenty of articles discussing the Claude Fable and its powers. I believe this article is interesting because I spend each day working with Claude Code and have been doing it basically every day since the beginning of the year 2026. When Claude Fable was released, I immediately tested the model to get to know it fully. I had a series of jobs that the Opus could one-shot or finish easily, which I tested Claude Fable on to really test its capabilities.
You should just check out this article to read my opinion on the capabilities of the Claude Fable, especially when comparing it to other borderline models like the Claude Opus 4.8 and GPT-5.5.
Why do you care about Claude Fable 5
First of all, we should start with why you should care about Claude Fable 5. This is probably the most anticipated LLM ever, as it had a lot of hype surrounding it for months before it was launched.
Anthropic itself spent a lot of time analyzing the model, talking about its capabilities and how dangerous it can be in the wrong hands.
So many people were excited about the launch of the model, and it was finally launched last week with full access to anyone who subscribes to Claude Pro or Max.
The model is made widely available to everyone, for everyone. Personally, I encountered no problems using the model until Saturday morning, Norwegian time. Even though the model was not available for a long time (I believe about 72 hours), I feel that I should test the model more and make a good impression on both its highs, lows, and overall strength.
What Claude Fable does well
First, I want to cover what Claude Fable did well. My overall opinion is that it is much better than Claude Opus 4.8. I have read some people on the internet saying how they feel the skills are not as good as the Claude Opus 4.8. In my opinion, this was clearly not true. I believe the people reporting this didn't test Claude Fable on complex enough tasks.
Of course, if you try Claude Fable on an already simple task that Claude Opus can do, you won't see its full potential. Where Claude Fable really shines is when you use it for more complex coding tasks.
Completing complex end-to-end tasks
I had several projects I was working on where I spent time with Claude Opus 4.8 to use it. Opus was able to use it, but it wasn't a one-shot launch, and I had to guide Claude Opus through another launch.
These activities were an example:
- A feature implementation where I had to change code in multiple repos
- Fixing an issue/bug a customer encountered, where I needed to improve LLM's information extraction capabilities
Unfortunately, I cannot elaborate on these issues because the work is in a closed codebase. What I did to compare Claude Opus and Claude Fable is that I had used Claude Opus before for these problems and solved them by guiding them step by step. Then I used it on Claude Fable, and it was able to shoot problems. A clear sign that Fable is a stronger model than Opus.
As a general note of Fable abilities, I would say the following:
Claude Fable is better able to complete tasks end-to-end, with a better understanding of both the ambiguity and the user's intent, better at implementing a real planned solution, and better at verifying that the solution is really correct by navigating around the computer or using integration tests.
I simply discovered that Claude Fable was willing to run for long periods of time, completing complex tasks without stopping or ending up in ongoing stories. I simply felt that tasks were now being done more autonomously, and I didn't have to provide as much guidance to make sure the model fit my goals.
Finding problems in codebases
Another amazing skill I saw in Claude Fable was that it was much better than Claude Opus in finding problems in codebases, or finding bugs or looking for refactoring opportunities or potential problems in the future.
I often use information like the one below to find problems in my codebase.
Scan thoroughly through the codebase to identify any potential bugs,
issues, or refactoring opportunities, and come back to me with an
HTML report with issues prioritized from most severe to least severe. With Claude Opus, I used this same instruction and did not get good results. Claude Opus couldn't find any refactoring opportunities or bugs, or the problems it found weren't really working. (Of course, note that this was after I had been doing a lot of refactoring and bug-finding with Claude Opus in a particular repository.)
However, when I ran Claude Fable just as quickly, it started to find a lot of serious problems, both security-wise and practical bugs, and also found a lot of refactoring opportunities that Claude Opus couldn't see.
I immediately started going through all the archives with Claude Fable using this information and fixing all the problems. I have pushed a lot of codes which has greatly increased the quality of my codes.
I think this is probably the single clearest sign that Fable is a stronger model than Opus. You can use the same information in the same codebase where Fable was able to find a bunch of problems that Claude Opus couldn't find.
I'm just glad that I was able to implement these refactorings, bug fixes, and more fixes before the model was unfortunately discontinued.
That's what Claude Fable doesn't do well
In the previous section, I covered what Claude Fable does best. I think it is also important to highlight the negatives of Claude Fable, given that it is not a perfect model.
Price
Claude Fable is definitely the most powerful coding model I've ever used. However, one of the biggest issues is how many tokens it spends to complete tasks.
Naturally, this is not a problem with the model itself. It's more of a problem with the rate limitations you have with Anthropic. Using Claude Fable with a subscription, I started hitting the subscription limit quickly.
This is quite a limitation, as you can no longer use the model indefinitely. In addition, I would argue that the subscription price of Claude Fable is too prohibitive for almost all companies. Using a model that costs $10 per million and $50 per million is not feasible for anyone but the largest companies.
Of course, you might argue that you can use Claude Fable only for programming and debugging, and then use Claude Opus for actual use. I agree that this can probably be done, and you will still get a lot of benefits from Claude Fable; however, spending a lot of time configuring which model to use is time consuming and definitely something you want to avoid if you want to be as productive and efficient as possible.
This is one of the worst, I would argue, limitations of rates and how much you can use the model/model costs when using API rates.
Sometimes the model is too ambitious
One small negative I would like to cover about Claude Fable is that the model is sometimes overly eager to find problems or make use of it. Sometimes I found that the model used things in a very complicated way. For example, changing more lines of code than was really necessary, or finding more problems in the codebase when most problems weren't that serious.
I find this a bit annoying at times, but I also believe this is a trade off that Anthropic has accepted. You, of course, want the model to always look for problems and always try to fix them, and you want the fixes to work quickly, of course. It is difficult to balance this eagerness and, at the same time, avoid the model from being overly eager to find problems and fix them.
All in all, though, this is a pretty minor disadvantage. It's just one little thing I noticed when using Claude Fable. Worst of all is the cost of the model.
The conclusion
In this article, I have compiled my thoughts on Claude Fable. I compared it to the previous frontier model from Anthropic, which was the Claude Opus 4.8. Claude's Fable 5 model is amazing, but it also has some disadvantages:
- One is that it is overly ambitious, which I believe is quite a bit.
- Prices. The price, of course, is an important part. However, it is not a problem with the model itself, and the best models will always cost more money. People can think for themselves if the model is worth what you have to pay.
Overall, though, it's a very powerful model. I hope that it will be available again and that other models, both from other frontier labs and open source models, will reach these capabilities in a few months, so that we have even more powerful coding agents for software engineering tasks.
👋 Touch
👉 My Free eBook and Webinar:
🚀 10x Your Engineering with LLMs (Free 3-Day Email Course)
📚 Get my free ebook Vision Language Models
💻 My webinar on Vision Language Models
👉 Find me on social media:
💌 Stack
🐦 X / Twitter
