Anthropic has launched Claude Fable 5, a model in the Mythos class adapted for public use. At the same time, the company released Mythos 5, which is built on the same architecture but comes with some safeguards partially removed and is available only to cybersecurity specialists.
The company stated that Fable 5 delivers better results on most benchmarks tested, and that its lead over Anthropic’s other models grows as tasks become more complex. For example, Mythos 5 reportedly accelerated parts of the drug development process by about ten times.
“A model with this level of capability carries risks, which is why Fable 5 automatically routes some requests to the less powerful Claude Opus 4.8,” Anthropic said.
The company cited cybersecurity, biology, and chemistry as examples of such requests, along with attempts at distillation — the process of extracting a model’s capabilities to train competing systems.