Claude Fable 5 reportedly routes risky topics to another model
Anthropic released on June 9, 2026, according to the item. It is described as the company’s strongest model, with top results across benchmarks for software engineering, research, and vision tasks. Its lead over older models is said to grow when tasks are longer and more complex.
The main change is its safety setup. When a question falls into a narrow set of risky areas, such as , biology, chemistry, or , does not answer directly. Opus 4.8 automatically takes over instead.
The user is always told when this happens, and it is said to affect less than 5% of sessions. A less restricted version called also exists, but it is currently limited to selected cyber defenders and critical operators in the Glasswing program. Anthropic plans to widen access through a trusted access program for defensive and biomedical research.
Key points
- is presented as a new Claude model released on June 9, 2026.
- It is described as strongest on long, complex software, research, and vision tasks.
- Risky , biology, chemistry, and questions are routed to Opus 4.8.
- is described as the same model without those limits, but only for selected trusted users.
- Developers should check official details before relying on this behavior in real products.