Anthropic's Claude Fable 5.1 Slashes Agentic Work Costs by 45% While Doubling Research Benchmarks

Anthropic has launched Claude Fable 5.1, slashing agentic work costs by up to 45% and more than doubling its predecessor's scientific research benchmarks. The new model, generally available, also introduces built-in watermarking and tops the Artificial Analysis Intelligence Index, while Claude Mythos 5.1, a less restricted version, is available for trusted access programs in cybersecurity and life sciences.

New Models and Availability

The launch includes two distinct configurations: Claude Fable 5.1 and Claude Mythos 5.1. Both models share the same underlying architecture but differ in their safety protocols. Fable 5.1 is designed for broad use and is generally available across major cloud platforms, including AWS, Google Cloud, and Microsoft Azure. Mythos 5.1, conversely, is tailored for high-stakes domains such as cybersecurity and life sciences, where its less restrictive safety filters are deemed appropriate. Access to Mythos 5.1 is currently limited to participants in Anthropic's Project Glasswing trusted access program.

Significant Cost Reductions for Agentic Tasks

A key update with Fable 5.1 is its revised pricing structure, particularly beneficial for agentic workflows. Anthropic estimates that Fable 5.1 will cost approximately 25 percent less than Fable 5 for standard workloads. For tasks that involve extensive agentic operations, the cost reduction can reach up to 45 percent. This is primarily driven by a reduction in cache read prices, which have decreased from $1 to $0.25 per million tokens. Input and output prices for Fable 5.1 remain consistent with previous models, at $10 per million tokens for input and $50 per million tokens for output.

Enhanced Performance in Coding and Research

The new models demonstrate notable performance gains in critical benchmarks. On the Terminal-Bench-Science 0.1 for agentic scientific research, Fable 5.1 achieved a score of 52.6 percent, more than doubling Fable 5's previous score of 24.7 percent. For agentic coding tasks, Fable 5.1 scored 55.8 percent on Terminal-Bench 4.0, while Mythos 5.1, with its specialized configuration, reached 60.9 percent. These scores represent a substantial improvement over Fable 5's 42.0 percent and GPT-5.6 Sol's 37.3 percent on the same benchmark.

Fable 5.1 also leads the Artificial Analysis Intelligence Index with a score of 66, surpassing Claude Opus 5 (63) and GPT-5.6 Sol (61).

New Safety Features and Enterprise Safeguards

Anthropic has integrated built-in watermarking into Fable 5.1 and Mythos 5.1, marking these as the first Claude models to include this feature. A detection API for these watermarks is currently in private preview. Additionally, Anthropic is rolling out Enterprise Frontier Safeguards (EFS) in phases starting this fall. EFS aims to provide zero-data-retention privacy for enterprise customers by enabling data storage within customer-controlled cloud infrastructure, enhancing data security and compliance.

The safety filters in the new models have also been refined, resulting in 60 percent fewer false positives in cybersecurity contexts and 85 percent fewer in biology-related applications. This improvement is particularly relevant for Mythos 5.1's targeted use cases.

Conclusion

The release of Claude Fable 5.1 and Mythos 5.1 represents a strategic update from Anthropic, focusing on both performance enhancements and cost efficiency for advanced AI applications. The significant price reductions for agentic work, coupled with improved benchmark scores in critical domains like scientific research and coding, position these models for broader adoption. The introduction of built-in watermarking and the upcoming Enterprise Frontier Safeguards also underscore Anthropic's commitment to responsible AI development and enterprise-grade security. Users interested in exploring the capabilities and pricing of these new models can find them generally available on major cloud platforms.

Sources

• Introducing Claude Fable 5.1 and Claude Mythos 5.1 \ Anthropic \ Anthropic
• Claude Fable 5.1 & Claude Mythos 5.1 System Card
• Anthropic launches Claude Fable 5.1 and says it’s up to 45 percent cheaper for agentic work | The Verge
• Anthropic's new Fable release is cheaper, less restrictive | TechCrunch
• Anthropic's Claude Fable 5.1 promises better coding and research at up to 45 percent less