AI research company Anthropic has unveiled Claude Fable 5.1 and Claude Mythos 5.1, positioning them as leading models for coding, knowledge work, and scientific research.
Fable 5.1 is available broadly across Amazon Web Services, Google Cloud, Microsoft Azure, and the Claude API, while Mythos 5.1 remains accessible only through verified access programs tailored for cybersecurity and life sciences professionals. The two models share an identical base architecture but operate under different safeguard configurations, with Mythos offering more permissive controls for vetted defensive research.
The release delivers measurable performance gains across agentic coding, multidisciplinary reasoning, and computer-use benchmarks. Early testing indicates Fable 5.1 resolves complex software issues more reliably than its predecessor, with partner organizations noting particular strength in long-horizon tasks and root-cause analysis. The model also demonstrates improved efficiency, completing demanding workflows with fewer tokens than earlier flagship versions. Beyond conventional workloads, the models show emerging scientific utility.
Mythos 5.1 designed high-affinity protein binders with hit rates approaching fifty percent across twelve targets, outperforming existing competition designs by an order of magnitude in binding affinity. Separately, Fable 5.1 produced a high-resolution elevation map of Venus using decades-old NASA radar data and optimized open-source genomics models through custom GPU kernels, reducing inference costs by up to sixty percent on genome-wide analyses.
Anthropic has restructured the economic and privacy framework surrounding the new models. Cache read pricing has been reduced by seventy-five percent to $0.25 per million tokens, a change that the company estimates lowers typical workloads by roughly twenty-five percent and highly agentic tasks by up to forty-five percent. Standard token pricing remains unchanged at $10 per million input tokens and $50 per million output tokens.
To address enterprise data concerns, Anthropic introduced Enterprise Frontier Safeguards, a system developed with input from over one hundred customers across regulated industries. EFS routes activity data to customer-controlled cloud infrastructure rather than Anthropic servers, providing privacy equivalent to zero data retention while preserving automated misuse detection across sessions and accounts. Human review, when necessary, is conducted by the customer rather than Anthropic.
The rollout begins this autumn; eligible customers may use interim zero data retention on Fable 5.1 and Fable 5 until EFS becomes available. The company also refined its biology and cybersecurity safeguards to reduce benign-content flagging by sixty and eighty-five percent respectively in key categories, and implemented invisible watermarking on outputs to comply with the EU AI Act. Distillation protections have been strengthened as well, including restrictions on manual context editing for new API accounts.
The post Anthropic Launches Dual Claude 5.1 Models With Advanced Scientific Capabilities And Zero-Retention-Equivalent Privacy For Regulated Industries appeared first on Metaverse Post.
