Comparing Fable and 10 other LLMs on refactoring a LangGraph god node

For what it is worth, your site has been blocked by the UK's national cyber security centre on my current connection:

    This site may be associated with malicious activity or malware.
     Access to this site has been blocked by the Protective DNS Service
    Site: wtf.korridzy.com
    Please contact your local Network Administrator or IT support if you require further assistance


    Look up this site on Cloudflare Radar [1] for more information.

[1] https://radar.cloudflare.com/domains/domain/wtf.korridzy.com

Takeaways Which model to use for generating architecture.

The simpler the decision you need to make, the more readily you can just take a proposal from Fable or GPT-5.5. And now that Fable is unavailable outside the US, the choice between Opus and GPT is far from obvious. As a quick default I'd lean toward GPT.

I had a pretty rough experience with Fable today. At first, Claude showed a message that Fable was available again, but if I typed /model it showed the same "disabled" state for the export controls, so that was sloppy. After I got it working (I had to login again), I asked it to do a security audit of a server I've written, and it spent a few tokens and then "flagged" the request and downgraded to Opus 4.8. Opus 4.8 just kind of glazed me for writing secure code without finding anything important.

By the way, I ran ccusage afterwards and if I had been paying API rates I would have apparently spent $6.88 for that.

Amazing stuff Anthropic.

Any chance you'll retest with latest iterations of models?

Opus 4.8, GLM-5.2, Kimi K2.7, etc

Wait what? gpt5.5 is better than fable. I thought fable was the endtimes?!

Wonder if this is because fable swirches down to opus for somethings without telling you?

Even for a ui change fable complained about policy issues. This model is unusable