Anthropic just rolled out Claude Fable 5.1 and Mythos 5.1—basically the same core AI model, but with different guardrails. Fable is open to everyone, while Mythos is only for vetted pros and organizations.

Performance and Cyber Tasks

In benchmarks, Fable 5.1 outperformed Fable 5, Claude Opus 5, and GPT-5.6 Sol when it came to coding, science problems, and automation. Mythos 5.1 pushes even harder, especially for cybersecurity use cases.

Security and Usage Rules

Fable 5.1 can now be used for software vulnerability hunting. Creating exploits, pentesting, and some other risky stuff are still off-limits. Anthropic says they’ve cut false cyber protection triggers by about 60%—before, even legit requests from security pros would get bumped to weaker models. You can check the updated policy and model behavior on the release page.

Pricing and DeFi Risks

It’s cheaper now: everyday tasks run about 25% less than Fable 5, and heavy-duty autonomous jobs are up to 45% cheaper. Some experts have warned that Mythos-level models could spike the number of hacks, since bug hunting is now faster and cheaper—a real risk for DeFi.