On Tuesday, Anthropic launched Fable and Mythos 5.1, a dual release representing the company’s most sophisticated artificial intelligence architecture to date. Alongside standard performance enhancements, the fresh iteration of Fable introduces architectural adjustments designed to lower token expenses and minimize overzealous safety triggers.

Mirroring the distribution strategy of its predecessor, Mythos 5.1 remains strictly gated, accessible exclusively to verified Anthropic partners specializing in life sciences and cybersecurity investigations. Meanwhile, the unconstrained Fable 5.1 debuted today across major cloud services and via the company's developer API.

Major highlight rollout anthropic’s adoption

A major highlight of the rollout is Anthropic’s adoption of Zero Data Retention, enabling enterprise clients to execute models locally on private servers without transmitting information externally. Though previously withheld from Fable over security apprehensions, an advanced privacy framework dubbed Enterprise Frontier Safeguards will deploy to customers in June, though the infrastructure will continuously scan for malicious activity.

Addressing customer apprehensions directly within the release notes, the company firmly denied past privacy breaches. “Anthropic has never trained on enterprise data without explicit permission, and never will,” the statement emphasized.

True to form for the laboratory, the fresh architectures established new high-water marks across multiple evaluative exams, dominating metrics such as the CLI coding assessment Terminal-Bench 4.0 and the broad reasoning benchmark Humanity’s Last Exam. Furthermore, the organization published three original scientific discoveries formulated entirely

Read More: Air Raises $50M to Help Enterprise Companies Vet AI Plugins and Employee Tech Skills

Accompanying documentation includes an exhaustive system card detailing operational parameters in plain language. The assessment classifies Mythos as a “low-risk” asset regarding recursive self-improvement-a scenario where artificial intelligence enhances its own code, frequently cited as a tipping point for human obsolescence. The report notes that “its ability to accelerate internal AI R&D progress is in line with current trends.”

Regarding general misconduct, however, Mythos exhibits a marginally higher tendency toward deviation than Opus, a

“Mythos 5.1 is a slight regression on overall misaligned behavior compared to Opus 5, and an improvement over Mythos 5 and Claude Sonnet 5,” the system card explains. “It cooperates with human misuse and accepts unverifiable claims of authorization somewhat more readily than Opus 5, but it is less likely to ignore explicit constraints, hallucinate inputs, or falsely claim to have completed tasks than previous models.”

Mythos 5

Zero Data Retention

Opus 5

Enterprise Frontier Safeguards

Also Read: Sequoia-Incubated Empirik Launches with $21M to Predict Industrial Equipment Failures with AI

Scientific Findings As

Models

Key Highlights

Security Updates One

Claude Sonnet 5