Anthropic Releases Claude Fable 5.1 and Mythos 5.1, Outperforming Predecessors on Multiple Metrics and Cutting Cache Read Fees by 75%

TradingKey - On September 1, Eastern Time, Anthropic announced the launch of its next-generation AI models, Claude Fable 5.1 and Claude Mythos 5.1. The company positioned these two models as "the world's most advanced models for coding and knowledge work."

Both models share the same infrastructure, with the key difference lying in their safety guardrail settings. Fable 5.1 is now officially available to individual and enterprise users, while Mythos 5.1 is accessible only through a trusted access program to vetted individuals and organizations in cybersecurity and life sciences.

Anthropic stated that the tiered design aims to balance capability deployment with access restrictions in high-risk scenarios. Both models feature a knowledge cutoff of June 2026, five months later than the previous generation.

In terms of performance, Fable 5.1 outperformed its predecessor Fable 5, Opus 5, and OpenAI's GPT-5.6 Sol across multiple benchmarks. On the agentic scientific research benchmark Terminal-Bench-Science 0.1, Fable 5.1 scored 52.6%, more than double that of Fable 5, while surpassing Opus 5's 29.0% and GPT-5.6 Sol's 22.4%.

[Source: Anthropic]

On the agentic coding benchmark Terminal-Bench 4.0, Fable 5.1 scored 55.8%, while Mythos 5.1, featuring fewer safety constraints, reached 60.9%. In knowledge work, Fable 5.1 also outperformed previous generations with a score of 1,853 on the GDPval-AA v2 test.

[Source: Anthropic]

Pricing adjustment is another key focus of this update. Cache read fees for Fable 5.1 were reduced by 75% to $0.25 per million tokens. Base pricing remains unchanged: $10 per million tokens for input and $50 per million tokens for output.

Anthropic noted that under typical workloads, the overall cost of Fable 5.1 is approximately 25% lower than that of Fable 5, while under highly agentic workloads, costs can be reduced by up to 45%.

[Source: Anthropic]

Significant improvements were also made in data security. Anthropic introduced a new architecture called Enterprise Frontier Safeguards (EFS), which allows customers to store data in their own controlled cloud infrastructure. The system will be rolled out to enterprise customers in phases this fall.

Before EFS becomes available, eligible customers can run Fable 5.1 under a zero data retention policy. Anthropic also pledged that it has never used, and will not use, enterprise data to train models without explicit permission.

Safety guardrail mechanisms have also been optimized. Anthropic stated that Fable 5.1 reduces cybersecurity false-positive blocks by approximately 60% compared to before. The model can now be used to identify software vulnerabilities, but will not be used to develop exploits targeting those vulnerabilities.

Fable 5.1 is now fully live and available across platforms such as the Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. Anthropic stated that Fable 5.1 will be retired no earlier than September 1, 2027.