Anthropic has launched Claude Fable 5.2 and Claude Mythos 5.2, two variants of its latest frontier AI model designed for technical research, complex coding, and agentic workflows. Fable 5.2 includes updated safety classifiers that restrict high-risk cybersecurity and biological research queries, making it broadly available across public platforms. Mythos 5.2 shares the same base model architecture but omits those specific filters, limiting access to approved organizations participating in Anthropic’s Cyber and Life Sciences Verification Programs.
According to benchmarks released by Anthropic, Fable 5.2 outperforms its Opus 5.5 model on several technical evaluations at significantly lower operational costs. When tested on coding and technical reasoning tasks, Fable 5.2 matched or exceeded prior top-tier benchmarks while operating up to 30 percent faster than 5.1 generation models. Anthropic also introduced an optional fast mode capable of increasing generation speed up to 2.5 times.
Fable 5.2 is priced at $8 per million input tokens and $40 per million output tokens on the Claude Platform, AWS, Google Cloud, and Microsoft Azure. Anthropic reported that safety interventions during training helped reduce false-positive refusals by 55 percent compared to previous iterations, while intentionally capping specialized autonomous AI research capabilities at Opus 5.5 levels.
Why it matters
Presents a dual-release model strategy that separates general enterprise deployments from high-risk bio/cyber research applications.
Dramatically reduces inference costs for complex agentic workflows, matching previous flagship intelligence at lower price points.
Provides fast-mode API options to support real-time developer applications and autonomous coding agents.
Source: marktechpost.com



