Anthropic Cuts Claude 5.2 API Prices 20% as Coding Scores Jump to 49.1%
Updated
Updated · InfoWorld · Nov 10
Anthropic Cuts Claude 5.2 API Prices 20% as Coding Scores Jump to 49.1%
3 articles · Updated · InfoWorld · Nov 10
Summary
Anthropic priced Claude Fable 5.2 and Mythos 5.2 at $8 per million input tokens and $40 per million output tokens, down from $10 and $50, while saying the models run up to 30% faster.
Fable 5.2 posted 49.1% on the FrontierCode 2.0 benchmark at Xhigh effort versus 32.4% for Fable 5.1, and Anthropic said its High effort mode can match Opus 5.5 coding performance at roughly half the cost per task.
Mythos 5.2, which shares the same base model but has tighter safeguards, rediscovered 89.6% of vulnerabilities in Anthropic's internal CVE test versus 78.3% for Mythos 5.1; direct access is limited to vetted cybersecurity and life-sciences groups.
An internal eight-day test had Fable 5.2 coordinate more than 1,000 subagents to build a bootable prototype operating system, but analysts said reliability, human oversight and integration across enterprise tools still limit broader autonomous deployment.
Data handling remains a hurdle for regulated customers: zero-retention users can keep those terms, while others face 30-day retention unless they use Anthropic's private-beta Enterprise Frontier Safeguards to keep safety monitoring within their own infrastructure.