.
Anthropic has introduced Claude Mythos 5.1, a specialised version of its latest AI model designed for advanced cybersecurity and life-sciences work, but the company is keeping access tightly controlled because of the model’s capabilities.
Mythos 5.1 was launched on September 1, 2026, alongside Claude Fable 5.1, Anthropic’s more broadly available model for coding, research and knowledge work.
Although the two models share the same underlying technology, Anthropic said they differ mainly in the safeguards applied to them. Fable 5.1 is generally available, while Mythos 5.1 is being offered only through trusted-access programmes for vetted professionals and organisations.
Mythos 5.1 Targets Cybersecurity and Life Sciences
Anthropic describes Mythos 5.1 as a model built for specialised work where conventional AI safeguards can restrict legitimate professional use.
The model is being made available to approved cybersecurity professionals through Anthropic’s Cyber Verification Program (CVP). A separate Life Sciences Verification Program (LSVP) is being developed for researchers who need access to its advanced capabilities in scientific and biological research.
Anthropic said the life-sciences programme has already enrolled its first participants in partnership with the US government, with plans to expand access to a broader community of professionals.
For cybersecurity, the company said the updated safeguards are designed to distinguish legitimate defensive activity from potentially harmful applications.
No Public Pricing for Mythos 5.1
Unlike Fable 5.1, Anthropic has not published a separate public price for Mythos 5.1.
The model is not offered as a standard consumer product or general-purpose API option. Instead, access is currently tied to Anthropic’s trusted programmes and approved organisations.
Anthropic said Mythos 5.1 is currently available only to a set of US organisations, although it is working with the US government to extend access to additional partners in the United States and internationally.
The restricted approach reflects Anthropic’s assessment that some of the model’s capabilities require additional oversight before wider deployment.
Same Core Model, Different Safeguards
Anthropic said Mythos 5.1 and Fable 5.1 are essentially the same underlying model.
The distinction comes from their safety configurations. Mythos provides more permissive safeguards to vetted users whose professional work would otherwise be restricted by the tighter protections applied to Fable.
The company said this approach is intended to give legitimate researchers and cybersecurity professionals access to advanced capabilities while maintaining restrictions against harmful applications.
Anthropic has also integrated Mythos 5.1 into Claude Security, its service for scanning codebases for vulnerabilities and suggesting patches for human review.
Advanced Scientific Performance
Anthropic’s testing suggests Mythos 5.1 can perform sophisticated scientific tasks.
In one experiment, the model was given access to protein-design and folding tools and asked to produce protein binders for 12 targets. Anthropic said nearly half of the designs were viable binders, significantly above the typical 10–15 per cent hit rate it cited for current protein-design work.
The company also reported that, for three targets, Mythos-generated binders recorded binding affinities up to 10 times stronger than the best designs submitted to certain protein-design competitions.
In another test involving computational biology, Mythos 5.1 created custom GPU kernels and cached intermediate results to accelerate seven open-source deep-learning models by as much as 2.5 times, while maintaining identical outputs.
Fable 5.1 Gets the Broader Release
While Mythos remains restricted, Fable 5.1 is the version Anthropic is putting into the hands of the wider market.
The company says Fable 5.1 improves on Fable 5 across coding, research, knowledge work and long-running tasks.
Its standard API pricing remains $10 per million input tokens and $50 per million output tokens, but Anthropic has reduced the price of cache reads by 75 per cent to $0.25 per million tokens.
The company estimates that the pricing change could lower costs by about 25 per cent for typical workloads, with savings reaching approximately 45 per cent for highly agentic, context-heavy tasks.
Fable 5.1 is available through Anthropic’s platforms as well as Amazon Web Services, Google Cloud and Microsoft Azure.
Anthropic Tightens Focus on Controlled AI Deployment
The release of Mythos 5.1 highlights Anthropic’s increasingly segmented approach to deploying powerful AI systems.
Rather than making every capability available to every user, the company is separating general-purpose AI from specialised systems that may require professional verification and additional safeguards.
Anthropic said it has also strengthened protections against AI capability distillation, a process through which advanced model capabilities can potentially be extracted and reproduced in another system.
For cybersecurity, the company has loosened restrictions on defensive vulnerability discovery while retaining limitations around activities such as penetration testing and exploit generation.
For now, therefore, Claude Mythos 5.1 is not a model ordinary users can simply subscribe to or purchase. Its deployment is being treated as a controlled programme for vetted cyberdefenders and life-sciences professionals.
The strategy reflects the growing challenge facing frontier AI companies: making increasingly capable models useful to legitimate researchers and businesses while limiting access to capabilities that could create significant security risks if deployed without adequate oversight.
