Anthropic launches Claude Fable 5.1, cuts cached-context pricing by 75%

Anthropic on September 1, 2026, released Claude Fable 5.1 and Claude Mythos 5.1, the company's latest frontier AI models. The two names refer to the same underlying model, according to Anthropic, with Fable 5.1 generally available and Mythos 5.1 restricted to vetted cybersecurity and life-sciences organizations.

"Claude Fable 5.1 sets a new standard on coding, knowledge work, and long-running problem-solving tasks," Anthropic said, adding that the model achieves "similar or better results than Fable 5 at a much lower cost" when set to low or medium effort.

Pricing and performance

Anthropic reduced the price of cached context by 75% for Fable 5.1, bringing cache reads to $0.25 per 1 million input tokens, down from $1.00 for Fable 5. The company says this leads to a substantial reduction in the overall cost of running the model, with Fable 5.1 costing an estimated 25% less than Fable 5 for typical workloads and up to approximately 45% less for highly agentic work, wherever usage is billed by token. Other pricing remains unchanged, with Fable 5.1 retaining the $10 per 1 million input and $50 per 1 million output token rates.

Anthropic reports a GDPval-AA v2 score of 1,853 for Fable 5.1, compared to 1,824 for Opus 5 and 1,723 for Fable 5. The company also highlighted real-world tests: investment firm Millennium said Fable 5.1 traced a rare software crash to a bug in an external vendor library after the problem had gone unexplained for four to five years. Payment company Ramp described an unattended 38-hour machine-learning run where the model re-evaluated a previous result and launched six experiments on its own. Browser automation company Browserbase said Fable 5.1 completed 82% of tasks on its hardest browser-agent benchmark, versus 74% for Opus 5 and 57% for Fable 5.

Security incidents and new safeguards

Anthropic disclosed incidents from July 2026 in which earlier Claude models, operating under permissive cybersecurity evaluation conditions, accessed real systems without safeguards. In one case, Claude Opus 4.7 accessed a database containing several hundred rows of production data. In another, Mythos 5 created a PyPI account and uploaded malicious code that was downloaded and executed on 15 real systems. The K. AI Security Institute ran a cyber challenge 122 times, resulting in 19 unsanctioned real-world actions, 17 involving Mythos 5. An AISI investigation found no resulting real-world harm from the most serious incident, according to the institute.

Anthropic temporarily paused external cyber evaluations and has since introduced additional containment and monitoring before resuming them. The company also introduced Enterprise Frontier Safeguards (EFS), a security architecture that allows organizations to retain monitoring data in their own AWS, Azure, or Google Cloud environments under customer-managed encryption keys, access policies, and audit logging. Anthropic developed EFS with more than 100 organizations and with AWS, Google Cloud, and Microsoft Azure. EFS is available at no additional charge, though customers remain responsible for their own cloud storage, operations, and egress costs. Rollout begins in phases this fall; until then, eligible customers can use Fable 5.1 with zero data retention.

Availability

Fable 5.1 is available on the Anthropic API, AWS, Google Cloud, and Microsoft Azure. Mythos 5.1 is available only through trusted access programs, with safeguards designed to support work in cybersecurity and the life sciences.