Anthropic Launches Claude Fable 5.1 and Mythos 5.1
Anthropic says the new models improve coding and knowledge work while lowering cache-read costs.
Anthropic has launched Claude Fable 5.1 and Claude Mythos 5.1, positioning the new models for coding, knowledge work, and complex long-running tasks. Claude Fable 5.1 is available broadly, while Mythos 5.1 is being offered through trusted access programs.
The biggest change is a lower cost for repeated context. Anthropic says cache reads with Claude Fable 5.1 cost 75% less than with Fable 5, reducing typical token-based costs by about 25% and by up to 45% for highly agentic workloads.
Claude Fable 5.1 targets long-running work
Anthropic says Claude Fable 5.1 sets a new benchmark for its coding and knowledge-work models and can match or exceed Fable 5 at lower effort levels. The company highlights complex, long-running tasks and research work as key uses.
Anthropic also says its cybersecurity safeguards now flag benign requests about 60% less often. On basic biology and medical questions, it reports an approximately 85% reduction in fallback rates.
Mythos 5.1 remains restricted
Claude Mythos 5.1 uses the same new model generation for higher-risk research and cybersecurity work, but Anthropic is limiting access to trusted programs rather than making it broadly available.
The launch also introduces Enterprise Frontier Safeguards, which Anthropic says will give enterprise customers stronger privacy protections while maintaining safeguards against adversarial use. The company says the rollout will begin in phases this fall.
Saganote has also covered Claude Opus 5 vs Fable 5, including the previous generation's position against Anthropic's other flagship models.