Anthropic's new Fable release is cheaper, less restrictive
News Source
β’Tue, 01 Sep 2026 19:39:22 +0000
π° What Happened
Anthropic released two new versions of its most advanced AI model on Tuesday: Fable and Mythos 5.1. The new Fable release is cheaper to use and blocks fewer harmless requests. Mythos 5.1 is only for approved partners working in cybersecurity or life sciences.
Fable 5.1, the unrestricted version, is available now through cloud platforms and the Anthropic API. The models set records on coding and reasoning tests. Anthropic also shared three scientific findings the models produced, including a map of Venus built from existing photos.
π The Backstory
Anthropic has been pushing for Zero Data Retention, which lets clients run its models without data leaving their own systems. A high-privacy service called Enterprise Frontier Safeguards will roll out to users in June. Clients will control how the monitoring happens.
Anthropic insists it never trained on enterprise data without permission, and says it never will. This release continues Anthropic's pattern of pairing a public model with a restricted research version. Each launch also comes with a detailed system card explaining the model's abilities.
π― Why It Matters
Cheaper, safer AI models mean businesses can use them without giving up customer data. If rivals follow, everyday AI tools may get better and cost less for everyone.
Anthropic released two new versions of its most advanced AI model on Tuesday: Fable and Mythos 5.1. The new Fable release is cheaper to use and blocks fewer harmless requests. Mythos 5.1 is only for approved partners working in cybersecurity or life sciences.
Fable 5.1, the unrestricted version, is available now through cloud platforms and the Anthropic API. The models set records on coding and reasoning tests. Anthropic also shared three scientific findings the models produced, including a map of Venus built from existing photos.
Anthropic has been pushing for Zero Data Retention, which lets clients run its models without data leaving their own systems. A high-privacy service called Enterprise Frontier Safeguards will roll out to users in June. Clients will control how the monitoring happens.
Anthropic insists it never trained on enterprise data without permission, and says it never will. This release continues Anthropic's pattern of pairing a public model with a restricted research version. Each launch also comes with a detailed system card explaining the model's abilities.
Cheaper, safer AI models mean businesses can use them without giving up customer data. If rivals follow, everyday AI tools may get better and cost less for everyone.