Back Issues/Search Home → Calendar → Archive → RSS → Subscribe → Services → Vendors → Blog → Current Issue → Popular → Newsletters (LinkedIn, Substack)

All issues › Volume 344, Issue 2 › IT News › AI

Claude Fable 5.2 vs. Fable 5.1: Fewer Refusals, with Small Savings on the Login Fix

The New Stack, Saturday, November 14th, 2026

In tests, Claude Fable 5.2 completed webhook and security tasks that Fable 5.1 refused, but its price cut saved little.

Jessica Wachtel compared Anthropic's Claude Fable 5.2 with Fable 5.1 on three developer tasks through the API, one run per model.

Fable 5.1 refused 34 of 50 turns in a webhook relay chat and both turns of a signature checker task, while Fable 5.2 completed both, passing all 50 hidden tests on the signature checker.

On a login library repair task, the models tied at 50 of 50 on all three versions.

Fable 5.2 produced about 30% more output tokens per second, but total time was about the same, and its estimated cost was only about 2% lower despite a 20% cut in per-token rates.

more →  ·  More from AI →