Anthropic has released Claude Fable 5.1 and Claude Mythos 5.1 as its new flagship models for coding and knowledge work, positioning them as the world's most advanced models for autonomous, multi-step tasks.
- The models achieve a new state of the art on benchmarks including Terminal-Bench-Science (52.6%) and HLE (59.1%), with an Artificial Analysis Intelligence Index score of 66.
- Anthropic introduced Enterprise Frontier Safeguards (EFS) for agent observability and zero-data-retention support to facilitate enterprise adoption.
- While input and output pricing remains unchanged, the cache read price was cut by 75% to $0.25 per million tokens.
- Early analysis suggests Fable and Mythos 5.1 may share underlying weights with different safety or routing behaviors rather than being distinct base models.
The release aims to address previous criticisms of impracticality by improving speed, verbosity, and tone, targeting users who require reliable autonomous agents for long-horizon problem solving.