Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Monday, the same day its IPO roadshow began. The company says the models are a step above Opus 5.5 in almost all capabilities it evaluated. Artificial Analysis places Fable 5.2 first on its Intelligence Index v5 with a score of 65 at max effort, ahead of Gemini 4 Carbon at 60 and Opus 5.5 at 59. Pricing is $8 per million input tokens and $40 per million output tokens, a 20% cut from Fable 5.1's $10 and $50. The lower sticker price does not translate directly into lower bills. SemiAnalysis's GDPx benchmark of professional tasks found the cost per task rose to $88.15 from $67.16 for Fable 5.1. Run logs show Fable 5.2 often wrote hundreds or thousands of test cases and launched subagents to check its own work, where Opus 5.5 and Fable 5.1 tried about 20. That extra checking helped it reach first place on GDPx at 16.4%. Anthropic reports that Fable at High effort matched Opus 5.5 at Max on FrontierCode 2.0 for about half the cost per task, though these are vendor figures. Medium is the default effort in the Claude apps and Claude Code. Fable 5.2 also requires data retention by default so Anthropic's safety classifiers can run, with zero data retention limited to eligible enterprise customers. Subscription access is restricted, with Fable capped at 50% of a weekly limit and tenure-based resets for Ultra and Team Premium users usable until March 8. For Australian businesses and professionals, the practical change is that a cheaper token rate does not guarantee a cheaper workflow, so cost should be measured per completed task at the effort level you choose. Testing Medium, High and Max on your own non-sensitive work before setting defaults is a sensible first step. Default data retention matters if prompts contain personal information, so check it against your privacy policy and the Privacy Act 1988 (Cth), including the cross-border disclosure rules in APP 8, and confirm whether your provider qualifies for zero retention. If you only use AI tools through a finished product and do not send data to the API, the practical impact this week is limited to whether the product you use switches to the new model.