Anthropic Launches Dual Claude 5.1 Models With Advanced Scientific Capabilities And Zero-Retention-Equivalent Privacy For Regulated Industries
AI analysis firm Anthropic has unveiled Claude Fable 5.1 and Claude Mythos 5.1, positioning them as main fashions for coding, data work, and scientific analysis.
Fable 5.1 is obtainable broadly throughout Amazon Web Services, Google Cloud, Microsoft Azure, and the Claude API, whereas Mythos 5.1 stays accessible solely by way of verified entry packages tailor-made for cybersecurity and life sciences professionals. The two fashions share an an identical base structure however function below completely different safeguard configurations, with Mythos providing extra permissive controls for vetted defensive analysis.
The launch delivers measurable efficiency beneficial properties throughout agentic coding, multidisciplinary reasoning, and computer-use benchmarks. Early testing signifies Fable 5.1 resolves advanced software program points extra reliably than its predecessor, with companion organizations noting explicit energy in long-horizon duties and root-cause evaluation. The mannequin additionally demonstrates improved effectivity, finishing demanding workflows with fewer tokens than earlier flagship variations. Beyond typical workloads, the fashions present rising scientific utility.
Mythos 5.1 designed high-affinity protein binders with hit charges approaching fifty p.c throughout twelve targets, outperforming current competitors designs by an order of magnitude in binding affinity. Separately, Fable 5.1 produced a high-resolution elevation map of Venus utilizing decades-old NASA radar knowledge and optimized open-source genomics fashions by way of customized GPU kernels, lowering inference prices by as much as sixty p.c on genome-wide analyses.
Enterprise Economics and Privacy Architecture
Anthropic has restructured the financial and privateness framework surrounding the brand new fashions. Cache learn pricing has been decreased by seventy-five p.c to $0.25 per million tokens, a change that the corporate estimates lowers typical workloads by roughly twenty-five p.c and extremely agentic duties by as much as forty-five p.c. Standard token pricing stays unchanged at $10 per million enter tokens and $50 per million output tokens.
To deal with enterprise knowledge considerations, Anthropic launched Enterprise Frontier Safeguards, a system developed with enter from over 100 prospects throughout regulated industries. EFS routes exercise knowledge to customer-controlled cloud infrastructure slightly than Anthropic servers, offering privateness equal to zero knowledge retention whereas preserving automated misuse detection throughout periods and accounts. Human assessment, when mandatory, is performed by the shopper slightly than Anthropic.
The rollout begins this autumn; eligible prospects could use interim zero knowledge retention on Fable 5.1 and Fable 5 till EFS turns into accessible. The firm additionally refined its biology and cybersecurity safeguards to scale back benign-content flagging by sixty and eighty-five p.c respectively in key classes, and carried out invisible watermarking on outputs to adjust to the EU AI Act. Distillation protections have been strengthened as effectively, together with restrictions on handbook context modifying for brand spanking new API accounts.
The submit Anthropic Launches Dual Claude 5.1 Models With Advanced Scientific Capabilities And Zero-Retention-Equivalent Privacy For Regulated Industries appeared first on Metaverse Post.
