According to an official release published by Anthropic on November 9, 2026, the lab has introduced Claude Fable 5.2 alongside Claude Mythos 5.2, upgrading its flagship reasoning models engineered to resolve highly complex mathematical, scientific, and software engineering challenges.
Both models share the same underlying architecture but diverge in deployment safety parameters. Claude Fable 5.2 incorporates hardened biological and cybersecurity guardrails to enable open access for commercial developers, while Claude Mythos 5.2 is reserved for vetted organizations operating within Anthropic's Cyber Verification and Life Sciences programs.
Independent benchmarking by Artificial Analysis placed Claude Fable 5.2 at the top of its global leaderboard, scoring 65 on the Artificial Analysis Intelligence Index—surpassing all 74 evaluated frontier models against an industry median of 40. The evaluation logged an output throughput of 103 tokens per second, with an average execution duration of 44.1 minutes and $26.54 in compute cost per comprehensive reasoning task.
During internal evaluation cycles, Anthropic reported that the models executed large-scale software migrations that previously stalled engineering teams, designed three functional electronic hardware devices from initial schematics, and resolved two open theoretical problems in black hole thermodynamics. Furthermore, the model formalized a complete machine-checked proof of Fermat's Last Theorem within a 30-hour reasoning window.
Standard API access is priced at $8.00 per million input tokens and $40.00 per million output tokens, marking a cost reduction from the $10/$50 rates of version 5.1. Anthropic also debuted a dedicated fast mode that accelerates token emission by up to 2.5 times at $16.00 input and $80.00 output per million tokens.