Anthropic Launches Claude Fable 5.1 and Mythos 5.1 With Advanced Coding and Research Capabilities

Anthropic has introduced Claude Fable 5.1 and Claude Mythos 5.1, two new AI models designed for complex coding, knowledge work and scientific research. While both models use the same underlying system, Anthropic has equipped them with different safeguards. Fable 5.1 is generally available, while Mythos 5.1 is restricted to vetted users working in areas such as cybersecurity and life sciences.

Fable 5.1 targets complex coding and knowledge work

Claude Fable 5.1 is positioned as Anthropic’s latest model for long-running and agentic tasks, with improvements over Fable 5 in coding, research and broader knowledge work.

Anthropic says typical workloads using Fable 5.1 are expected to cost around 25% less than Fable 5, while highly agentic workloads could be up to 45% cheaper. The reduction comes primarily from a major cut in cache-read pricing. Cache reads will cost $0.25 per million tokens, while standard input and output pricing remains at $10 and $50 per million tokens respectively.

The model also posted stronger results on several internal benchmarks. It scored 52.6% on Terminal-Bench-Science 0.1, compared with 24.7% for Fable 5 and 29% for Opus 5. On CursorBench 3.2.0, which evaluates agentic coding, Fable 5.1 scored 73.4%, ahead of Fable 5 at 70.5%.

Anthropic also highlighted a real-world example involving investment firm Millennium, where Fable 5.1 reportedly identified the cause of a rare software crash that engineers had been unable to explain for several years. The model traced the issue to a bug in an external vendor library.

AI models take on scientific research

Anthropic is also highlighting the models’ ability to assist with scientific work. In one experiment, Claude Fable 5.1 trained a neural network to produce a higher-resolution elevation map covering about one-third of Venus using radar imagery from NASA’s Magellan mission.

The resulting map can show surface details down to around 2–3 kilometres, compared with 10–20 kilometres in the earlier map. Anthropic said the new map also represents heights up to 25% more accurately.

Mythos 5.1 demonstrated capabilities in molecular design as well. Anthropic gave the model access to open-source protein-design and folding tools and sent its designs for experimental validation. Across 12 targets, nearly 50% of the designs were confirmed as viable protein binders, while typical hit rates in protein design are around 10–15%, according to Anthropic. On three targets, some of its binding affinities were reported to be 10 times higher than the best designs submitted to Adaptyv Bio’s protein-design competitions.

The model was also used to optimise seven open-source deep-learning models by writing custom GPU kernels and caching intermediate results. Anthropic said these optimisations delivered speed improvements of up to 2.5 times while producing identical outputs, with estimated GPU cost reductions of 30–60% for certain large-scale analyses.

Mythos 5.1 remains restricted

Despite sharing the same underlying model, Mythos 5.1 is not being released for general use. Anthropic says it is available through trusted-access programmes for vetted cybersecurity professionals and life-sciences researchers.

The restricted approach reflects the higher risks associated with advanced capabilities in these areas. Anthropic’s testing found Mythos 5.1 to have the strongest cybersecurity capabilities of any model it has released, while still placing it in the lower risk category under its own framework.

The company has also introduced more precise safeguards for Fable 5.1. Its cybersecurity protections now allow the model to identify software vulnerabilities for defensive purposes, although activities such as penetration testing, exploit generation and binary-based vulnerability scanning remain restricted. Anthropic says Claude Code users can expect around 60% fewer cybersecurity safeguard interventions per session compared with the previous safeguards used with Fable 5.

Anthropic expands access while maintaining safeguards

Fable 5.1 is available across Anthropic’s platforms, including its API and major cloud services. Mythos 5.1 is currently limited to selected US organisations and vetted professionals, with Anthropic working to expand access to additional partners.

The company is also strengthening privacy and enterprise controls through its Enterprise Frontier Safeguards, which are designed to let customers keep data within their own cloud infrastructure while maintaining protections against misuse. The system is scheduled to roll out in phases beginning this fall.

The launch reflects Anthropic’s broader strategy of making increasingly capable AI systems useful for demanding professional and scientific tasks while applying tighter controls to areas where the same capabilities could create significant security risks.

Conclusion

Claude Fable 5.1 and Mythos 5.1 mark another step in Anthropic’s push toward AI systems that can work on extended technical, scientific and research problems rather than simply answer individual prompts. Fable 5.1 brings these improvements to general users, particularly in coding and knowledge work, while Mythos 5.1 keeps its more sensitive cybersecurity and life-sciences capabilities behind restricted access.

With stronger benchmark performance, lower costs and more targeted safeguards, the two models show how frontier AI development is increasingly focused not only on improving raw model intelligence but also on determining how those capabilities can be deployed safely in the real world.