Anthropic on November 9 released Claude Fable 5.2 and its restricted counterpart Claude Mythos 5.2, positioning them as tools for complex coding, reasoning, research, and vision tasks. Both models share the same underlying system, with Fable adding stricter safeguards for cybersecurity and biology, while Mythos is reserved for verified organizations in those fields.
Pricing for the new models is $8 per million input tokens and $40 per million output tokens, down from $10 and $50 in the previous generation. Anthropic says output is up to 30% faster than Fable 5.1 and Mythos 5.1. An optional fast mode can generate output up to 2.5 times faster, costing $16 per million input tokens and $80 per million output tokens; for Mythos, fast mode is limited to organizations in verification programs.
Benchmark results shared by Anthropic include a 49.1% score on FrontierCode 2.0 for Fable 5.2, which the company calls the highest of any released model it has tested. At High effort it matched Opus 5.5's best score of about 45% at roughly half the cost per task. Fable 5.2 scored 64.9% on RoboDojo Sim, compared with 29.8% for GPT-6 Astra and 27.6% for Fable 5.1. On Humanity's Last Exam with tools it scored 74.2%. Mythos 5.2 found 89.6% of vulnerabilities on the CVE Rediscovery test, 10.2 percentage points above Opus 5.5 and 14.4 points above Mythos 5.1.
Real-world tests underscored both capability and cost. AI YouTuber Matthew Berman described Fable 5.2 as the best model he has used, while tracking estimated API expenses for ambitious projects. A Gears of War-inspired level took about 43 hours and 48 minutes, processed 8.63 billion tokens, and carried an estimated cost of $14,710. An Excel clone ran up an estimated $11,700 after cached-input discounts, and a Fall Guys-style game consumed more than 9 billion tokens over 71 hours for roughly $13,000. Anthropic also reported that more than 90% of changes merged into its core codebase are now written by Claude with human guidance. A team of 1,000 Fable 5.2 agents formalized the proof of Fermat's Last Theorem in 30 hours, and the model built a prototype operating system that boots on a laptop after coordinating over 1,000 subagents for eight days.
Safety measures were also updated. Anthropic said Fable 5.2 produced about 55% fewer blocks than Fable 5.1, with cybersecurity blocks down about 60% and biology blocks down about 40%. The company said its anti-distillation protections cut data harvested by distillers by 90% to 99%. Paid subscribers on eligible plans get temporary Fable access or usage resets, but heavy workloads can exhaust allowances quickly; Anthropic noted that it does not accept kidneys as payment.
