Anthropic launches Claude Fable 5.1 and restricted Mythos 5.1 for advanced research

The two models share the same underlying capabilities, but Anthropic is limiting Mythos 5.1 to vetted cybersecurity and life sciences users while making Fable 5.1 generally available

Anthropic has released Claude Fable 5.1 for general use while restricting the same model's more permissive Mythos 5.1 configuration to vetted cybersecurity and life sciences researchers

The two models share the same underlying capabilities, but Anthropic is limiting Mythos 5.1 to vetted cybersecurity and life sciences users while making Fable 5.1 generally available

Anthropic has released Claude Fable 5.1 alongside Claude Mythos 5.1, pairing a generally available frontier model with a more permissive version that is restricted to approved cybersecurity and life sciences users.

"Claude Fable 5.1 and Claude Mythos 5.1 are the same model, but with different levels of safeguards," Anthropic states. Fable 5.1 is available across Claude's platforms, while Mythos 5.1 can access capabilities that would otherwise trigger restrictions when used for specialist biological or cybersecurity research.

That separation reflects how Anthropic is handling increasingly capable models. Rather than making the full research capability available by default, it is keeping higher-risk functionality behind verification programs while trying to reduce unnecessary blocks for mainstream users.

Fable 5.1 also changes the economics of using Claude for long-running AI work. Anthropic has cut cache-read pricing by 75 percent to $0.25 per million tokens. Cache reads allow the model to reuse context it has already processed rather than processing the same material again.

The company estimates that this will reduce typical workload costs by about 25 percent compared with Fable 5, rising to around 45 percent for complex coding and highly agentic tasks where large amounts of previous context are repeatedly reused. Standard input and output prices remain $10 and $50 per million tokens respectively.

Anthropic pushes Claude further into scientific research

Some of the more consequential claims around the new models concern scientific research rather than conversational AI.

In Anthropic's testing, Claude Mythos 5.1 designed protein binders that were subsequently sent to two external organizations for experimental validation. Across 12 targets, Anthropic says almost half of the designs were viable binders. On three specific targets, the company reports binding affinities 10 times higher than the best designs submitted to Adaptyv Bio's protein design competitions.

Those results should be read as Anthropic-reported research rather than a broad demonstration that the model can independently develop medicines. Designing a molecule that binds effectively to a biological target is an early part of drug development, not evidence that a treatment will ultimately work.

Claude Fable 5.1 was also used to create a higher-resolution elevation map covering about one-third of Venus. The model worked from radar data captured by NASA's Magellan mission and an existing map covering one-fifth of the planet. Anthropic says the resulting map resolves details at two to three kilometers rather than 10 to 20 kilometers and improves height accuracy by up to 25 percent.

The map is being released under a Creative Commons license ahead of NASA's VERITAS and the European Space Agency's EnVision missions.

In computational biology, Mythos 5.1 modified seven open-source deep-learning models to run as much as 2.5 times faster while producing identical outputs. Anthropic estimates that the optimizations could reduce GPU costs by 30 to 60 percent in some genome-scale analyses and says it plans to open-source the work.

More capable models bring tighter access controls

The same scientific capabilities are one reason Mythos 5.1 is not being released on the same terms as Fable 5.1.

Anthropic says Mythos 5.1 has stronger biological and cybersecurity capabilities than its predecessor. Its evaluations found that the model remained below the next biological risk threshold in the company's Responsible Scaling Policy, while its cybersecurity capabilities stayed within the lower risk category of Anthropic's Frontier Compliance Framework.

Access to Mythos is therefore being handled through two programs. The Cyber Verification Program provides reduced safeguards for approved defensive security work, while the Life Sciences Verification Program is intended to give professional researchers access to advanced biological capabilities. Anthropic says the latter has been developed with the US government and has already enrolled its first participants.

Fable 5.1 is becoming less restrictive in some areas. It can now identify software vulnerabilities for defensive purposes, although exploit generation and several other dual-use security tasks remain restricted. Anthropic says its updated cybersecurity safeguards produce around 60 percent fewer interventions per Claude Code session than the safeguards used with Fable 5.

The company also reports that biology safeguards now trigger 85 percent less often on benign elementary biology and medical questions.

Enterprise customers will get another safeguard system, Enterprise Frontier Safeguards, that stores customer data within cloud infrastructure controlled by the customer rather than Anthropic. The system is due to roll out in phases beginning later this fall.

Fable 5.1 adds EU AI Act watermarking

The release also introduces a less visible change to Claude's outputs. Anthropic signed the European Union AI Act's Code of Practice on Transparency of AI-Generated Content in July. As a result, models released after August 2 must add a numerical watermark that can be used to assess whether Claude was likely involved in producing a piece of text.

The watermark is invisible in normal use and, according to Anthropic, does not contain information about the user, organization, or conversation.

A detection API is now entering private preview for eligible groups including regulators, law enforcement, media organizations, fact-checkers, researchers, educational organizations, and EU civil society groups.

Claude Fable 5.1 is available now through Claude, the Claude API, Amazon Web Services, Google Cloud, and Microsoft Azure. Claude Mythos 5.1 remains restricted to vetted cybersecurity and life sciences users, with Anthropic currently providing access to a set of US organizations while working to expand the program.