Fable 5 came back thinner. Turns out it was the bouncer, not the model. 🚪

Claude Fable 5's July 1 reinstatement triggered swift backlash on social media, with users calling the model "broken," "nerfed," and "lobotomized" in posts collected by BharadwajC (@bwjbuild) on July 2, 2026. Two benchmarks published data the same day and reached opposite conclusions: BridgeMind's BridgeBench AI reported severe quality degradation, while Arena.AI found differences so small they may not be relevant enough to notice. Both, in their own way, are correct—the model did not get dumber, but the classifier in front of it became far more aggressive.

BridgeMind re-ran its full coding suite against the July 1 version of Fable 5 the day it came back. Debugging scores fell from 86.2 to 25.9, Refactoring from 73.6 to 38.4, and Hallucination Resistance from 75.9 to 61.7 on BridgeBench's 0–100 scale, which tests real-world coding tasks including debugging, refactoring, and hallucination resistance. The drop, however, reflects routing rather than capability. Of 12 TypeScript debugging tasks, only three actually reached Fable 5; the remaining nine were intercepted by Anthropic's safety classifier and rerouted to Claude Opus 4.8, and BridgeBench scores every fallback as zero because the responding model was not the one under evaluation. BridgeMind posted the figures publicly on July 2.

The classifier was deployed as a condition of Fable 5's reinstatement and was trained to block the Amazon-reported jailbreak technique that had enabled the model to identify and demonstrate software vulnerabilities. It works as intended, and it also triggers on benign inputs—TypeScript debugging apparently resembles "security work" closely enough to fire the fallback on most attempts.

Arena.AI reached a different verdict by measuring perceived quality rather than routing paths. The platform aggregates thousands of blind human-preference votes across Text, Vision, Document, Code, and Agent categories and ranks models using Elo scoring, a chess-derived rating system that adjusts for statistical uncertainty across head-to-head matchups. Arena.AI stated that the community has been asking how Claude Fable 5 compares before vs. after its latest re-deployment and that the platform collected thousands of votes on the new endpoint across Arenas to produce its comparison, with margin gaps too narrow to register as meaningful degradation.