Anthropic has launched Claude Fable 5.1 and Claude Mythos 5.1, two new AI models designed for coding, research and demanding knowledge-based tasks. The company says the models can handle longer and more complex workloads while reducing costs for many users.

The two versions are based on essentially the same underlying model but operate with different safety controls. Claude Fable 5.1 is available broadly, while Mythos 5.1 is reserved for vetted users working in areas including cybersecurity and life sciences.

Anthropic described the models as its most advanced offerings for coding and knowledge work, while pointing to their research capabilities as an early indication of how AI could support scientific work.

Claude Fable 5.1 targets lower AI costs

One of the main changes with Claude Fable 5.1 is its pricing structure.

Anthropic says typical token-based workloads will cost around 25% less than with Fable 5. For tasks where the AI operates more autonomously, takes multiple steps and uses external tools, the company estimates savings of up to 45%.

The reduction is largely linked to a 75% cut in cache-read pricing. Cache reads allow the system to reuse information it has already processed instead of processing the same information again.

Fable 5.1 continues to cost $10 per million input tokens and $50 per million output tokens. Cache reads are priced at $0.25 per million tokens.

Anthropic also reports stronger performance on several coding tests. On CursorBench 3.2.0, Fable 5.1 recorded a score of 73.4%, compared with 70.5% for Fable 5 and 70% for Opus 5.

Anthropic highlights scientific research capabilities

The company is also positioning the new Claude models for research-intensive work.

In one demonstration, Claude Fable 5.1 was used to create a higher-resolution elevation map of part of Venus from older radar imagery collected during NASA's Magellan mission. Anthropic says the resulting map could show details measuring about two to three kilometres, compared with 10 to 20 kilometres in the earlier map.

Claude Mythos 5.1 was tested in protein design as well. Anthropic says the model generated protein binders that produced strong laboratory results. Across 12 targets, the reported hit rate was close to 50%, compared with a typical 10% to 15% rate cited by the company for protein design.

Mythos 5.1 was also used to optimise seven open-source biology and genomics models. Anthropic says those models ran up to 2.5 times faster, while estimated GPU costs for some genome-wide analyses dropped by 30% to 60%.

Mythos 5.1 remains restricted

Anthropic is not releasing Claude Mythos 5.1 to the general public.

Access is being provided through trusted programmes for selected cybersecurity professionals and life sciences researchers. The additional access is intended for users who require capabilities that remain restricted in the standard Fable 5.1 model.

For defensive cybersecurity work, Anthropic says Fable 5.1 can assist with identifying software vulnerabilities. Activities such as exploit generation and penetration testing remain restricted.

The company also says its updated cybersecurity safeguards generate about 60% fewer false-positive interventions than its previous safeguards.

For life sciences, Anthropic has established an access programme for professionals using Mythos 5.1 in research and development, with the programme developed in partnership with the US government.

Anthropic adds new enterprise data controls

Anthropic is also introducing Enterprise Frontier Safeguards, or EFS, for enterprise customers.

The system is designed to offer privacy comparable to a zero-data-retention setup while allowing Anthropic to retain safeguards intended to prevent misuse. Under the arrangement, customer data is stored in cloud infrastructure controlled by the customer rather than Anthropic.

Anthropic plans to begin rolling out EFS to enterprise customers in phases later this fall.

Claude Fable 5.1 availability

Claude Fable 5.1 is available immediately across Anthropic's platforms. Developers can use the model through the Claude API, while access is also available through cloud services including Amazon Web Services, Google Cloud and Microsoft Azure.

Mythos 5.1 remains limited to vetted users. Anthropic's approach reflects the different risk levels associated with giving broader access to advanced capabilities in cybersecurity and biological research.