Anthropic splits its new Claude model between open use and trusted access
Anthropic has introduced Claude Fable 5.1 and Claude Mythos 5.1, describing them as the same underlying model with different safeguard settings. Fable 5.1 is generally available, while Mythos 5.1 is limited to trusted access programs for cybersecurity and life sciences work. The release also changes token economics, enterprise data handling and how some sensitive research capabilities are made available.Availability and token pricing
Anthropic says Claude Fable 5.1 is available now across its own products and major cloud platforms, including Amazon Web Services, Google Cloud and Microsoft Azure. Developers can use the model on the Claude API under the model name claude-fable-5-1, while Anthropic says the broader Claude product surface also supports the release.The clearest commercial change is cache read pricing. Anthropic says cache reads, where the model reuses context it has already processed, now cost 75% less at $0.25 per million tokens wherever usage is billed by token. Other Fable 5.1 pricing remains the same as Fable 5: $10 per million input tokens and $50 per million output tokens.
Anthropic estimates that these changes reduce typical workload costs by about 25% compared with Fable 5. For complex coding and highly agentic tasks, where cache reads make up more of the workload, it says savings could reach about 45%. These are company estimates rather than independently verified market prices, but they show where Anthropic expects the economics of long-context and tool-heavy work to improve.
Enterprise safeguards and data retention
The release introduces Enterprise Frontier Safeguards, or EFS, a system Anthropic says is intended to combine misuse detection with enterprise privacy equivalent to zero data retention. Instead of storing customer data on Anthropic systems, EFS stores data in cloud infrastructure controlled by the customer.Anthropic says human review is, by default, handled by the customer rather than Anthropic. The company says it developed EFS with more than 100 customers across sectors including financial services, healthcare, manufacturing, telecom, law, retail and the public sector, along with cloud partners Amazon Web Services, Google Cloud and Microsoft Azure.
EFS is planned to roll out in phases beginning later this fall. Until it is available, Anthropic says eligible customers can use Fable 5.1, and Fable 5, with zero data retention. The practical implication is that Anthropic is trying to address one of the main enterprise objections to frontier AI deployment: how to monitor risky use without requiring vendors to retain sensitive operational data.
Mythos 5.1 is the restricted version of the same model
Anthropic describes Claude Mythos 5.1 as identical to Fable 5.1 at the model level, but with more permissive safeguards for vetted users whose work is constrained by cybersecurity and life sciences restrictions. Access is not general: it is routed through trusted programs.For cyberdefense, Anthropic says its Cyber Verification Program currently provides access to certain Opus and Sonnet-class models with reduced cyber safeguards and will add Mythos-class access in the near future. For life sciences, the Life Sciences Verification Program is designed to let professionals use Mythos 5.1 with safeguards tuned for professional research and development while other safeguards remain in place.
The company says it has enrolled first participants in the life sciences program in partnership with the US government and plans to expand access to the broader life sciences community. Anthropic also says Claude Security, its product for scanning codebases for vulnerabilities and suggesting patches for human review, is now powered by Claude Mythos 5.1.
Coding, knowledge work and scientific claims
Anthropic presents Fable 5.1 as a step up in coding, knowledge work and long-running problem-solving tasks, but the public evidence in the announcement is mostly company-reported benchmark discussion and customer testing comments. The article should therefore treat the performance claims as Anthropic’s account, not as independent validation.The company says Fable 5.1 can identify software vulnerabilities, but not develop exploits for them. Anthropic also says updated cybersecurity safeguards reduce false positives: in one formulation it says the newest safeguards block 60% fewer false positives than before, and elsewhere says Claude Code users can expect around 60% fewer safeguard interventions per session compared with previous Fable 5 safeguards.
Anthropic also describes early scientific work. It says Mythos 5.1 designed high-affinity binders that were experimentally validated by two external organizations, with binding affinities on three targets 10 times higher than the best designs submitted to Adaptyv Bio protein design competitions and a nearly 50% hit rate across 12 targets. For Venus mapping, Anthropic says Fable 5.1 trained a neural network to generate a high-resolution elevation map for one third of Venus, using NASA Magellan radar imagery and an existing map, with detail down to two to three kilometers rather than 10 to 20. Anthropic says it is releasing that map under a Creative Commons license ahead of NASA VERITAS and ESA EnVision missions.
Safety testing, safeguards and watermarking
Anthropic says it tested Fable 5.1 and Mythos 5.1 across chemical and biological risk, cyber risk, agentic safety and alignment before release. For chemical and biological risks, the company says Mythos 5.1 is more capable than Mythos 5 but remains below the next risk tier in its Responsible Scaling Policy, so it is deploying the model with the same safeguards applied to Mythos 5.For cyber risks, Anthropic says Mythos 5.1, tested with cybersecurity safeguards off, has the strongest cyber capabilities of any model it has released while still falling within the lower category of risk in its Frontier Compliance Framework. The company also says it commissioned external testing and automated testing by Gray Swan, and that it has not found evidence of a critical-severity jailbreak for Fable 5.1 safeguards. These are self-reported safety findings and should be read as part of Anthropic’s own release case.
The announcement also covers EU AI Act compliance. Anthropic says that in July 2026 it signed the Code of Practice on Transparency of AI-Generated Content, which required a watermark on outputs of models released after August 2, 2026. The company says the watermark is invisible without a detection API, does not affect output quality or content, and contains no user, organization or conversation information. A detection API is being rolled out in private preview for eligible organizations and certain enterprises with compliance obligations.
Conclusion
Claude Fable 5.1 is a product launch, a pricing change and a governance adjustment in one release. The lower cache read cost targets the economics of long-context work, while EFS is aimed at enterprise privacy concerns that often slow adoption.Mythos 5.1 is more sensitive. Anthropic is keeping it behind trusted access programs for cyberdefense and life sciences, even though it says the model is the same as Fable 5.1 underneath. That split is likely to become a recurring pattern for frontier AI releases: broader access for general capability, narrower access where the same capability touches dual-use research or security work.
Sources
Editorial Team - CoinBotLab
