Anthropic Launches Claude Fable 5.1 and Mythos 5.1 for AI Tasks
The artificial intelligence firm Anthropic unveiled Claude Fable 5.1 and Claude Mythos 5.1, claiming they are the most sophisticated models globally for coding and knowledge tasks.
Anthropic states that both offerings share the same foundational model architecture but include varying levels of operational safeguards. The company announced that Claude Fable 5.1 is broadly accessible to users, while Claude Mythos 5.1 is limited to trusted access programs that have safeguards set up specifically for cybersecurity and life sciences research.
"They represent the most sophisticated models globally for programming and knowledge tasks— and their research functionalities provide an initial look at how AI models can aid in scientific advancement," states Anthropic.
The company announced that Claude Fable 5.1 outperforms its predecessor, Fable 5, in handling long-duration problem-solving and technical activities. Anthropic observed that when functioning under Low or Medium effort settings, Fable 5.1 produces similar or better results at lower computational costs.
"Fable 5.1 is projected to be around 25% cheaper than Fable 5 for standard workloads, in cases where billing is done per token." "This is due to our lowering prices on cache reads," states Anthropic.
The organization also remarked that, "For highly agentic tasks, the savings can frequently be much greater--nearly 45%."
During practical assessments carried out by the investment company Millennium, the system pinpointed the cause of an occasional internal crash that human engineers and previous models had not been able to fix for years.
Also Read: Tech Leaders Join G20 Meeting as AI Regulation Debate Intensifies
Regarding data privacy, Anthropic presented a framework called Enterprise Frontier Safeguards (EFS), which keeps customer data within customer-controlled cloud environments instead of company-operated servers. EFS rollouts for enterprise accounts are set to begin in phases later this fall, featuring interim measures allowing zero data retention for qualifying users.
Also Read: Fervo Energy to Supply 396 MW Geothermal Power to Google
Regarding pricing, Anthropic announced that they have lowered the cost of cache reads for Fable 5.1 (where the model reuses previously processed context) in cases where usage is charged by token, like on their API. Cache reads now incur a cost that is 75 percent lower, at USD 0.25 per million tokens.