Anthropic released Claude Fable 5, its most powerful publicly available AI model, bridging the gap between frontier capability and safety. Following a restricted rollout through Project Glasswing, the company now offers Fable 5 to all users, while Claude Mythos 5 remains accessible only to vetted partners in cybersecurity and biology.
A Model for the Masses with Built-in Restrictions
Fable 5 shares the same core architecture as Mythos but includes automatic guardrails that detect high-risk requests. When a prompt involves cybersecurity, biology, chemistry, or model distillation, the system reroutes it to Claude Opus 4.8. Anthropic reports that over 95% of sessions run fully on Fable 5 without fallback. The model is priced at $10 per million input tokens and $50 per million output tokens, making it the most expensive general-purpose model on the market.
Sponsored Protocol
Record-Breaking Coding Performance
Benchmarks showcase Fable 5's strength. On SWE-bench Pro, it scores 80.3%, far ahead of GPT-5.5 at 58.6%. Stripe used Fable 5 to complete a codebase-wide migration in one day, a task that would have taken a team over two months manually. On FrontierCode Diamond, Fable 5 achieves 29.3% versus 13.4% for Claude Opus 4.8. Long-horizon autonomous coding is where the model excels, enabling enterprises to delegate complex software engineering projects.
Safety, Data Retention, and Enterprise Implications
Mythos 5 lifts some restrictions for trusted users, offering full capability in sensitive areas. Both models now require 30-day data retention, a policy Anthropic says is necessary for monitoring misuse. For regulated industries, this tradeoff between capability and data control will be a critical factor. The launch sets a precedent for how frontier AI labs can commercialize powerful models while gating dangerous capabilities. Enterprises must decide whether the gain in autonomous work justifies the new data governance requirements.
Sponsored Protocol
For more on cybersecurity risks, see our coverage of Critical Linux Bug and Tchap Breach.