After its June launch as Anthropic's top widely available model, Claude Fable 5 hit U.S. export limits and returned July 1 with stricter safety classifiers. Those rules now shunt more prompts—especially in sensitive areas—straight to the older Opus 4.8, which explains the measured collapse on BridgeBench debugging tasks from 86.2 percent down to 25.9 percent.
The Great Anthropic Bait And Switch.
Keep sending them your money!
FABLE 5 CAME BACK NERFED.
We re-ran the July 1st version of Claude Fable 5 on BridgeBench.
The results are brutal:
Debugging: 86.2 → 25.9 Refactoring: 73.6 → 38.4 Hallucination: 75.9 → 61.7
The new guardrails are kicking in on way too many tasks and falling back to Opus 4.8.
This is not the model that got banned.
Anthropic owes everyone an explanation.
Developers running bug diagnosis or code repair on TypeScript now see frequent hand-offs to the weaker model and explicit notifications when it happens. Early launch claims of fallback in under 5 percent of sessions no longer match current user reports.
Anthropic has said the post-redeployment classifiers will be tuned further, yet no dates or target trigger rates have been released. How quickly the original benchmark numbers might return therefore remains an open variable for teams counting on Fable 5.
Positive users praise Claude Fable 5's real-world usefulness on projects after Anthropic's guardrail updates, while negative users call the changes a lobotomy that killed the model and made it unusable.
No Digg Deeper questions have been answered for this story yet.
Claude Feeble 5 😮💨
Fable 5 isn't nerfed, it's SLAUGHTERED.
the problem isn't even the model itself, but the hard guardrails Anthropic has set in place.
seriously wtf anthropic? No wonder they were able to re-release Fable 5.
Fable 5 isn't nerfed, it's SLAUGHTERED.
the problem isn't even the model itself, but the hard guardrails Anthropic has set in place.
“We are sorry”—Anthropic
The constitutional AI fable has failed again.
Get used to my posting of how this failed Anthropic philosophy is a human extinction event if played out.
ouch!
Fable 5 isn't nerfed, it's SLAUGHTERED.
the problem isn't even the model itself, but the hard guardrails Anthropic has set in place.
Feels like an end of era, ordinary people will probably never again get upgraded frontier models.
Fable 5’s return shows how safety routing can downgrade a frontier model.
Now we only permissioned intelligence.
The cost of putting a gatekeeper inside intelligence.
To note, that safeguard is not a simple refusal layer; it is a classifier that sends flagged Fable 5 requests to Opus 4.8.
Fable 5 came back, but the old promise did not.
End of an era. ☹️
FABLE 5 CAME BACK NERFED.
We re-ran the July 1st version of Claude Fable 5 on BridgeBench.
The results are brutal:
Debugging: 86.2 → 25.9 Refactoring: 73.6 → 38.4 Hallucination: 75.9 → 61.7
The new guardrails are kicking in on way too many tasks and falling back to Opus 4.8.
This is not the model that got banned.
Anthropic owes everyone an explanation.
Maybe they didn't tell it to make all the efforts
FABLE 5 CAME BACK NERFED.
We re-ran the July 1st version of Claude Fable 5 on BridgeBench.
The results are brutal:
Debugging: 86.2 → 25.9 Refactoring: 73.6 → 38.4 Hallucination: 75.9 → 61.7
The new guardrails are kicking in on way too many tasks and falling back to Opus 4.8.
This is not the model that got banned.
Anthropic owes everyone an explanation.
in case I wasn’t clear, this is a routing problem not the model itself. the routing classifiers which Anthropic mentioned will improve, are redirecting some of the requests to Opus 4.8
Fable 5 isn't nerfed, it's SLAUGHTERED.
the problem isn't even the model itself, but the hard guardrails Anthropic has set in place.
FABLE 5 CAME BACK NERFED.
We re-ran the July 1st version of Claude Fable 5 on BridgeBench.
The results are brutal:
Debugging: 86.2 → 25.9 Refactoring: 73.6 → 38.4 Hallucination: 75.9 → 61.7
The new guardrails are kicking in on way too many tasks and falling back to Opus 4.8.
This is not the model that got banned.
Anthropic owes everyone an explanation.
@Hesamation what'd they do to my boy
@Ananth7e thats honestly disgusting isnt it
@Hesamation Fable 5 isn't what it used to be. I just sent them a project for audit, and it replied, “we can't do that; it violates our rules.” but there's nothing wrong with the project (it's my personal one)
@0x_kaize safeguards are harsh
@Hesamation Anthropic's safety guardrails worked so incredibly well, Fable 5 forgot how to debug entirely.
@kimmonismus well well well
@Hesamation Blame the government they are the ones who forced anthropic to put those guardrails 😒
@beffjezos Tbh it still feels a notch above Opus 4.8.
@Youssofal_ @Hesamation Industry standard 😂
@navigateny @Hesamation You're probably running opus 4.8 and didnt even realize. Most of the time it subtle changes to opus 4.8 and doesnt even tells you. Especially if using on terminal
@Hesamation Rip claude. Let give Chinese people chance
@SquidTakopi @Hesamation If I pick fabel in the model picker and it gives me the results based on those scores shown above it doesn't really matter if they're "Opus" scores. The reality is I just got a much shittier product
