Claude Fable 5 • Hacker News Board

Re: Claude Fable 5

#701

Post by mer_mer » Wed, Jun 10, 2026, 1:19 AM UTC

simonw wrote: post #46

Pelican for Fable 5 on default settings is a clear improvement on Opus 4.8 Fable 5 default: https://gist.github.com/simonw/036bee5a703e7ec84e34efa974438... Opus 4.8 (the "max" one is closest to Fable): https://simonwillison.net/2026/May/28/claude-opus-4-8/#and-s... Now here are the Fable pelicans for all five of the thinking effort levels - low, medium, high, xhigh, max: https://tools.simonwillison.net/markdown-svg-r…

It's interesting that Gemini 3(.1?) Deep Think is still the best at this task and it's still not really generally available. Maybe Fable could match it at higher effort levels? https://simonwillison.net/2026/Feb/12/gemini-3-deep-think/

Re: Claude Fable 5

#702

Post by tomjakubowski » Wed, Jun 10, 2026, 1:22 AM UTC

Paging senko, let's see Fable's oneshotted RTS!

https://senko.net/vibecode-bench/

Re: Claude Fable 5

#703

Post by solenoid0937 » Wed, Jun 10, 2026, 1:22 AM UTC

steve_adams_86 wrote: post #696

I'm using it to review recent work and it's doing a genuinely excellent job. This is a clear step up. Fewer decisions I have to guide it away from, faster conclusions on planning, more willing to go out of the way to make the correct decisions possible... This is really interesting. It feels like going from Sonnet to Opus, but, of course as a step up from Opus. This feels more like working with a competent peer than…

Why is your comment so grey/downvoted? One of the only actual usage experiences posted in this thread.

Re: Claude Fable 5

#704

Post by thepotatodude » Wed, Jun 10, 2026, 1:24 AM UTC

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

Organ segmentation with CNNs. Very disappointing.

Re: Claude Fable 5

#705

Post by tomjakubowski » Wed, Jun 10, 2026, 1:26 AM UTC

simonw wrote: post #534

I've spent enough time with this now in Claude Code (and Claude.ai and Claude Code for web) to have an opinion on Fable 5: it's a beast. I'm throwing some VERY difficult problems at at - things I've been dragging my heels on for months - and it's crunching through them very happily. One that I'm willing to share (albeit from just a week ago) - I built a Python library last week that bundles MicroPython compiled to WA…

What are some reasons to consider your project instead of Pyodide?

Re: Claude Fable 5

#706

Post by scotty79 » Wed, Jun 10, 2026, 1:33 AM UTC

Curiously nothing on DeepSWE and ARC-AGI-3 yet. For ARC at least there's a statement that Anthropic won't guarantee them that their secret private test data won't be collected by them and used for training.

Re: Claude Fable 5

#707

Post by chr15m » Wed, Jun 10, 2026, 1:38 AM UTC

I found this juxtaposition of facts telling:

> Drug design: Using Mythos 5, our internal protein design experts accelerated... Nine of the 14 protein targets from this study (shown below) yielded strong candidates for *drug design that we’re currently investigating*.

(emphasis mine)

> queries that are beneficial in the hands of cybersecurity professionals and biology researchers could be dangerous if available to malicious actors... When Fable’s classifiers detect a request related to cybersecurity, *biology and chemistry*, or distillation, the response is automatically handled by Claude Opus 4.8 instead.

All of the things they are nerfing are things that they also intend to profit from themselves.

- Cybersecurity - selling this to companies and US gov through "Glass Wing".

- Selling inference (distillation risk).

- And now, drug design.

I'm extrapolating "currently investigating" to "are going to monetize" but I don't think that's a big stretch. They appear to be using safety as a cover for anti-competitive behaviour.

Re: Claude Fable 5

#708

Post by elzbardico » Wed, Jun 10, 2026, 1:40 AM UTC

Anthropic sucks. but this paragraph should be in the "annals of AI-aided self-inflicted learned helplessnes":

> If Claude gives me poor or incorrect advice while I’m working on an AI component, I have no way of knowing whether the model was confused, whether my problem is unsolvable, or if some invisible policy restriction quietly kicked in.

Have you considered actually learning the theory, spending some time actually reading the papers and latest books, paying careful attention even to the eventual math here and there?

Re: Claude Fable 5

#709

Post by hankbond » Wed, Jun 10, 2026, 1:43 AM UTC

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.

Re: Claude Fable 5

#710

Post by vb-8448 » Wed, Jun 10, 2026, 1:43 AM UTC

On python coding is definitively better that everything else: clean and not overengineered code, understands very well the code base.

The only thing I'm wondering if they on purpose downgraded opus 4.8 performances in the last days before the release just to make the "step" look bigger. I'm pretty sure they did it also in the past with all other opus 4.x releases.