Megadose AI progress, ranked and analyzed.

Anthropic tests Fable 5.2 and Opus 5.5 ahead of the release

· TestingCatalog ·
Users are spotting unusually advanced Claude coding outputs that may be routed through unreleased Anthropic models.

TestingCatalog reports X posts showing alleged Fable 5.2 outputs, including one-shot JavaScript games and animated canvas scenes with no external assets or tools. Anthropic has not commented on the apparent A/B test. The piece also cites rumors that Opus 5.5 is being tested under `claude-wafer-eap`, with an announcement possibly coming this week. TestingCatalog's note

score 5

Categories: Model Releases