Updated: Anthropic introduces Claude Fable 5.1 and limited Mythos 5.1

Anthropic introduced Claude Fable 5.1 together with Claude Mythos 5.1 on September 1, 2026. According to the company, both use the same base model, but differ in their safety restrictions and access modes. Fable 5.1 is generally available, while Mythos 5.1 is intended for vetted organizations and individuals in trusted-access programs.

The new generation targets demanding programming, knowledge work, research and long-running agentic tasks. Alongside the release, Anthropic reduced the price of cache reads and adjusted its cyber and biological protection mechanisms.

Claude Fable 5.1: specifications, availability and pricing

Claude Fable 5.1 is available through the Claude API, Amazon Bedrock, Google Cloud and Microsoft Foundry. The model offers a context window of 1 million tokens and a maximum output of 128,000 tokens.

The price remains $10 per million input tokens and $50 per million output tokens. Cache reads cost $0.25 per million tokens, one-quarter of the previous price. The reduction is particularly relevant for applications that repeatedly send the model the same documents, instructions or other extensive context.

Anthropic estimates that, for typical tasks, cheaper cache reads could reduce costs by approximately 25 percent compared with Fable 5. For agentic workflows with intensive repeated context use, it cites savings of up to approximately 45 percent. These are the manufacturer’s estimates; the actual difference will depend on the share of cache reads and the architecture of the specific application.

Cyber protections are intended to produce fewer false positives

Anthropic claims that the new cyber protection mechanisms block 60 percent fewer harmless requests than the previous version. The model can assist in finding software vulnerabilities, but is not intended for exploit development.

The distinction is important for security teams using generative AI for defensive work. Fewer erroneous interventions should limit cases in which a protection system blocks legitimate analysis, while the boundary against creating exploitable code remains in place. The figure on the decrease in false positives comes from Anthropic and is not yet an independently verified measurement.

Enterprise Frontier Safeguards and data protection

Anthropic also announced Enterprise Frontier Safeguards. According to the company, the system is intended to store data in cloud infrastructure fully controlled by the customer and provide privacy protection comparable to a zero-data-retention mode. Rollout to enterprise customers is expected to take place gradually beginning in fall 2026.

Until the system becomes available, eligible customers are expected to have the option of using Fable 5.1 with zero data retention. The practical terms, scope of availability and independent assessment of the claimed properties remain issues to monitor.

Mythos 5.1 remains in a controlled-access mode

Claude Mythos 5.1 is not generally available. Anthropic offers it only to vetted experts and organizations in trusted-access programs, particularly for cyber defense and biological sciences. The company describes it as a variant with the same capabilities as Fable 5.1, but with protection mechanisms and access tailored to sensitive areas.

Claims about performance, safety properties and differences between the two variants come primarily from the manufacturer. After the release, it will therefore be important to follow independent benchmarks, customer experiences with actual costs and the results of external security testing.

Verified and updated: 09/02/2026 09:28