Anthropic launched Claude Fable 5.1 for general availability and Claude Mythos 5.1 for a restricted group of vetted cybersecurity and life sciences organizations on Tuesday, Sept. 1. The two products use the same underlying model but apply different safeguards, giving enterprise buyers a broader commercial option while preserving tighter access to the system’s most sensitive capabilities.
The release matters for companies choosing models for software development, research and other complex knowledge work. Anthropic is pairing claimed performance gains with lower effective pricing, fewer false-positive safety interventions and a new enterprise privacy architecture. TechCrunch independently confirmed the launch and availability structure.
Fable 5.1 targets cost and enterprise adoption
Anthropic said Fable 5.1 is available through its API, Claude products and major cloud platforms. The company estimates that typical token-billed workloads will cost about 25% less than with Fable 5 because it reduced pricing for cache reads, which reuse previously processed inputs. Anthropic said highly agentic workloads could see savings of as much as 45%, depending on usage patterns.
Those figures are company estimates, not guarantees for every deployment. For technology leaders, the relevant measure will be total cost per completed task, including retries, human review, integration work and the amount of context an agent repeatedly processes. A lower cache-read price could matter most for long-running coding and research systems that revisit large codebases or document collections.
Anthropic also said Fable 5.1 improves on its predecessor across internal and third-party benchmarks for coding, computer use, scientific research and business workflows. Benchmark claims should be treated as directional until customers reproduce them on their own workloads. The company’s launch materials include endorsements from early-access users, but those accounts came from organizations participating in prerelease testing.
One model, two safeguard levels
Mythos 5.1 remains limited to vetted organizations because Anthropic considers its cybersecurity and biology capabilities more sensitive. The company said the model is available through trusted-access programs for approved cyberdefenders and life sciences researchers, initially at a starting price of $10 per million input tokens and $50 per million output tokens.
Fable 5.1 applies additional restrictions to the same underlying model. Anthropic said its updated cyber safeguards produce about 60% fewer interventions per Claude Code session than the previous Fable safeguards. The model can help identify software vulnerabilities, but requests involving exploit generation, penetration testing and binary-based vulnerability scanning may still be redirected or blocked. Anthropic separately said its updated biology safeguards intervene on benign elementary biology and medical questions 85% less often than the safeguards introduced with Fable 5.
The split lets Anthropic sell a broadly available frontier model while reserving less-restricted capabilities for organizations that pass additional review. It also creates a procurement distinction: buyers are not only selecting a model, but accepting a particular safeguard and monitoring regime.
Privacy controls become part of the model decision
Anthropic announced Enterprise Frontier Safeguards, a system designed to provide misuse monitoring while keeping customer data in cloud infrastructure controlled by the customer. The company said the service will roll out in phases beginning this fall across Claude and cloud-partner environments. Eligible customers can use Fable 5.1 with zero data retention until the new system is available.
For chief information officers and security leaders, the operational questions now extend beyond benchmark scores. Teams will need to verify where prompts, outputs and monitoring records are stored; who can review flagged activity; how restrictions affect legitimate security work; and whether projected token savings survive real production workflows.
Anthropic’s release increases pressure on enterprise AI vendors to compete simultaneously on capability, task-level cost, privacy architecture and safety precision. The strongest procurement case will come from measured results inside a company’s own systems, backed by clear controls for data, access and human oversight.
Sources: Anthropic launch materials and product documentation; TechCrunch reporting.