Anthropic has released Claude Fable 5.1 and Claude Mythos 5.1 as the new top of its model lineup. The two are the same model with different levels of safeguards. Alongside gains in coding and long-running tasks, this release tackles three things customers complain about most: price, data retention, and safeguards.

Same Model, Different Safeguards

Fable 5.1 and Mythos 5.1 are identical models running under different safeguard settings. Fable 5.1 is generally available, while Mythos 5.1 is restricted to vetted users through trusted access programs. Mythos safeguards are designed specifically to support work in cybersecurity and the life sciences.

There are two routes to Mythos 5.1. The Cyber Verification Program (CVP) covers defensive security work, and the Life Sciences Verification Program (LSVP) covers life sciences professionals. The LSVP was built in partnership with the US government and has already enrolled its first participants. Claude Security, Anthropic's product that scans codebases for vulnerabilities and proposes patches for human review, now runs on Mythos 5.1 as well.

Lower Effort Settings That Still Match the Previous Generation

Fable 5.1 targets coding, knowledge work, and long-running problem solving. The interesting part of Anthropic's framing is the shape of the performance curve: at Low or Medium effort, the model matches or beats Fable 5, and at High effort it pulls well ahead. The defaults are High in Claude Code, and Medium in Claude Cowork and on Claude.ai.

Anthropic emphasizes that the model avoids shortcuts and digs into root causes. The example given comes from the investment firm Millennium, where Fable 5.1 identified the cause of a rare crash in internal systems that the firm's own engineers, and every other model, had failed to explain over several years.

The Price Cut Comes From Cache Reads

For typical token-billed workloads, Fable 5.1 costs an estimated 25 percent less than Fable 5. The saving comes from a reduction in cache-read pricing, which applies when the model re-reads inputs it has already processed and stored. Agentic workloads that revisit the same context repeatedly benefit the most, with savings reaching roughly 45 percent.

Data retention gets a new framework as well. Enterprise Frontier Safeguards (EFS) keeps data in cloud infrastructure the customer controls rather than Anthropic's own systems, and any human review is by default carried out by the customer. The aim is to deliver the privacy of a zero data retention agreement while still detecting misuse. EFS was designed with more than 100 customers across financial services, healthcare, manufacturing, telecom, law, retail, and the public sector, together with Amazon Web Services, Google Cloud, and Microsoft Azure. It rolls out in phases starting this fall, and eligible customers can use Fable 5.1 with zero data retention until it arrives.

Safeguards Tuned for Fewer False Positives

Rather than tightening restrictions, Anthropic tuned its safeguards for precision. In cybersecurity, the new safeguards block 60 percent fewer false positives than before. Part of the reason is that Fable 5.1 can now be used to discover software vulnerabilities. Developing exploits for them remains out of scope, and dual-use tasks such as penetration testing, exploit generation, and binary-based vulnerability scanning are still routed to Opus-class models.

In biology, safeguards now fire 85 percent less often on benign questions about elementary biology and medical topics compared with the versions that shipped with Fable 5. Queries that move into life sciences research and development are still directed to Opus models.

Protein Design and a New Map of Venus

The scientific results are what Anthropic is putting front and center this time.

In molecular design, Mythos 5.1 was given open-source protein design and folding tools, and its designs were sent to two external organizations for experimental validation. On three targets, its binding affinities were 10 times higher than the best designs submitted to Adaptyv Bio's protein design competitions, and its hit rate reached nearly 50 percent across 12 targets. Typical hit rates in protein design today sit at 10 to 15 percent, so this is roughly three to four times the norm.

For computational analysis, Fable 5.1 trained a neural network to build a new high-resolution elevation map covering a third of Venus, working from radar images taken by NASA's Magellan mission more than 30 years ago plus an existing map of one fifth of the planet. Resolution improves from 10 to 20 kilometers down to 2 to 3 kilometers, and heights are up to 25 percent more accurate. The map is being released under a Creative Commons license in the hope that it helps the upcoming NASA VERITAS and ESA EnVision missions choose observation targets.

In computational biology, Mythos 5.1 wrote custom GPU kernels and cached intermediate results, speeding up seven open-source deep learning models by as much as 2.5 times with identical outputs. Because these models are run thousands of times in a single experiment, estimated GPU costs drop by 30 to 60 percent. Work that would normally take a team of performance engineers weeks was done in days using publicly available source code alone, and Anthropic plans to open-source the optimizations soon.

Anti-Distillation and the EU AI Act

Fable 5.1 also strengthens defenses against distillation, the practice of extracting a model's capabilities. Specifically, API accounts created from the announcement date onward can no longer manually edit Claude's prior context in a multi-turn conversation while preserving the transcript of Claude's earlier thinking. That closes a publicly documented distillation technique. Existing accounts are not affected for now, but the change will apply to all users with future model releases.

On regulatory compliance, Anthropic signed the EU AI Act's Code of Practice on Transparency of AI-Generated Content in July 2026, alongside 190 other signatories. As a result, outputs from models released after August 2, 2026 carry a watermark that indicates numerically how likely it is that Claude wrote a given piece of text. The watermark is invisible without the detection API, has no impact on output quality or content, and carries no information about the user, their organization, or their conversations. The detection API is now in private preview for eligible organizations such as regulators, media, fact-checkers, researchers, and educational institutions.

Summary

Claude Fable 5.1 is an update that goes after operational friction, not just benchmarks: price, data retention, and false positives all get attention. The cache-read price cut in particular means agentic usage patterns see the biggest reduction on the actual bill. The scientific results, from protein binders to the Venus elevation map, are Anthropic's own figures, but external experimental validation and a Creative Commons map release put them in a form others can check.

Note: the thumbnail image is an AI-generated illustration.