Claude Fable 5

I got a content rejection for this question in a new chat. > What is the optimal EPA oil intake for nootropic effects? Very advanced classifiers they have.

HN needs pagination or sth alike - this page breaks my iPhone XS ;)

absolutely beast model but the token consumption is the 2x then the opus 4.8 what do you think about this ? i think that it should only use for the more complex task otherwise you have to run out of the limit..

The way the guerrilla marketing campaigns have been going on and IPOs left/right, I won't be surprised if GPT Next comes up and offers the same but unrestricted

Open source models seems to be 1-2 years behind the frontier, so I am very excited to see what happens when those open source labs get their hands on capabilities like this to accelerate their own development speed.

I uploaded to it my 23andme DNA test results and it refused to analyze it :(a

Oh my god it's actually here

Fable's ridiculous. It's flagging basic biology research questions as a security risk. I'm talking basic fundamental genetics topics that make working on any genetics-adjacent codebase unusable.

https://artificialanalysis.ai/evaluations/humanitys-last-exa... Not bad

its good for difficult problems, bad for design and code gen

On my very first Fable 5 prompt, got flagged on a hard but completely uncontroversial option math problem, many tokens in. Although it's pretty clear that this is an unremarkable experience at this point.

Anybody could suggest me how to use keep using Fable in claude code but with lesser rate limits? Any suggesstions?

I've seen enough degradation of the models I pay for from Anthropic to not bite. Fable will work fine for the first couple of weeks and then start degrading like previous models did.

I just tested it with a max subscription. On Ultracode mode, Fable 5 ate up 10% of my weekly allowance in 30 minutes. Granted, won't be using UC mode frequently, but still.

Careful using this with Cursor, especially for corp use. Anthropic will "retain agent request and output data associated with this model, regardless of you Cursor Privacy Mode setting."

Completely unusable for my usecase. Constant safety filters. Have not even been able to use it.

Organ segmentation with CNNs. Very disappointing.

Looks like a good model (sir). Costs are getting out of control though. 2x Opus and non-metered usage going away. We're quickly approaching the cost of a human salary for normal usage.

I’m sure this is banged on somewhere but I love their product branding, particularly how they have this “minor” “major” thing going on. Sonnet-Opus, and now Fable-Myth.

I’ll ask it to write me some win32 ui crap when I get hands on it, it will need all its brainpower to get that idiocy right.

I tried to give it something challenging but not something that is too much and it ate the entire session budget on this task alone.

Weird how every new model seems hyped up as the most dangerous yet and the one that will destroy society as we know it. They are also a commercial product.

Cursor users will note that the privacy setting and data retention is not the same as the other models.

Not sure I should use this for work just yet.

I use AI for a wide variety of things, of which technical is only a small part - and then it's usually a problem with project configuration, not coding. Why? Because I am often testing projects handed in by students. Projects that supposedly work on their machine, but certainly do not on mine.

Anyway, anecdotally, I find Copilot shockingly awful. It makes random changes to files that have nothing to do with the problem. Call it out, and it makes other changes to other irrelevant files.

ChatGPT and Gemini are both much better. Grok also isn't bad. Claude, I honestly haven't tried yet on these issues. Perhaps I should...

This thread takes >10s to load on my pc. Maybe after a certain number HN should fold comments? or a depth of >5?

Crazy and Scary! But its not for every one, you need to have a meaty thing for it to devourer and a deep enough pocket for it to devourer also.

I swear nowadays AI api pricing is getting to high like what the hell is 50 dollars for million tokens

If the claimed capabilities are true, Fable 5 is already at a superhuman level. We might see genuine unprecedented leaps in technology now, across all fields.

The OSS-Fuzz section is interesting. They compare it to their other models but carefully avoid comparing it to, you know. Fuzzing.

Ask Claude Code (I tried on Opus 4.8) to do this: "create a file with ISO country mappings"

API Error: Output blocked by content filtering policy

can't use it for code review

> Fable 5's safety measures flagged this message for cybersecurity or biology topics. They may flag safe, normal content as well. These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine them. Switched to Opus 4.8. Send feedback with /feedback or learn more

super

It's surprisingly sensitive to biology research topics - even reviewing standard papers on tissue culturing is flagged as a problem

It's getting harder to review the plans with Fable. So do we plan with Opus and let Fable implement or just start trusting blindly. Feels to me that this is another shift in how we operate these systems.

Does the model take some time to perform better?

Because I am running Opus and Fable side by side, Opus 4.8 is solving my coding problems better.

I just gave it a go at a problem I've been working on this week. Nothing fancy, just some inefficient code that we've been adding incremental improvements to for a while now to the point where some out-of-box thinking is probably required to push it any further – something Fable is obviously more than capable of.

After Fable did some thinking for a few minutes it gave some suggestions. A couple of them were valid – but very low impact, bordering on entirely pointless – but it's main suggestion.. It told me to make an update that would very clearly break the existing functionality.

So I thought about it for a moment...

Hm, I mean, I guess we could do that if we also did x, y & z to mitigate the behaviour change – maybe that's what Fable was thinking?

I replied, explaining that it would change the behaviour, assuming it would explain what it was thinking given there was clearly more to it. But no, it just said it was wrong.

This isn't some super advanced or complex code either. Had I gave this question to a senior engineer in a technical interview and they gave the answer Fable gave me I would view that very negatively. I was expecting something creative and interesting, not irrelevant + incorrect.

I'm sure it's a step up from 4.8 (although am not interested in burning the tokens to find out), but this clearly isn't as significant a change as some are implying. I'm sure if I asked it to come up with some out-of-box suggestions it could, but any competent engineer would have realised that by themselves.

I’m waiting to see results on deepswe - that benchmark really seemed accurate for opus and gpt 5.5…

pdf gives 404

is this a good time to hussle for my "AI does not need a break but you do!"* app? as quite a lot of people will propably get ai brain exhaustion maximising "playing" with that new model until they take it away again?

* https://rainbreak.franzai.com/

Cracks me up that a system “card” is 319 pages.

Seems like all a bad actor has to do to gain access is to compromise one of the partner companies that has access.

The safety filter is awful on this one.

It's more like a free trial, because the model is going to become pay-per-query in 10 days

Not comparing to GPT Pro models is a bit strange, considering that's the natural comparison

First shot's for free

How much and what kind of data do you need to throw at these models to get a good design interface?