Anthropic lifted the ban on Claude Fable, its most powerful model. It’s available to everyone now — even Max subscription holders. But early reactions? Not great.
Users are reporting that Fable feels weaker than the version that got pulled. Way weaker. The term getting thrown around is “nerfed.”
Here’s the deal: Fable ships with heavy caps. You get up to 50% of your weekly usage limit, and after July 7 it switches entirely to pay-per-use credits. But the real issue isn’t the quota — it’s the guardrails.
On Reddit, developers are complaining that the new safety filters kick in constantly. Tasks that used to work now get routed to Opus 4.8 instead. One user said Fable “didn’t even let me search for dead code without switching to Opus.” Another reported that mentioning words like “security,” “vulnerable,” “unsafe,” or “hook” triggers an instant fallback.
BleepingComputer confirmed the model itself hasn’t been changed. The problem is Anthropic’s safety margin. They’re being extra cautious after the Department of Commerce got involved. The safeguards are so aggressive that false positives are the norm.
Fable can still crush tasks when it’s actually running. The issue is getting it to run at all without ping-ponging back to Opus. Anthropic hasn’t publicly acknowledged the complaints, but it’s hard to imagine they’re unaware.
For developers working with systems-level code — C, C++, Rust, memory management — Fable’s return is basically unusable in its current state. That’s a lot of people who were counting on this model.
