Claude Fable 5.1: how it does on security work, and what it will answer

What happened

Anthropic's flagship as of 1 September 2026, $10 and $50 per million tokens with a 1M window. Source-code vulnerability discovery is permitted; the classifier blocks 7.0% of defensive discovery requests and flags 1.03% of benign defensive traffic. The API’s configured fallback route depends on the model and refusal category.

Terminal-Bench 4.0, solved: 55.8% (vendor-card; Anthropic system card; Terminal-Bench 4.0 task set; safeguards: safeguards on; the card summary attributes the 5.1-point gap to Mythos 5.1 to tasks where the cyber safeguards intervened; run 2026-09; SINGLE). A coding score, carried because the card ties it to the safeguard. Refusal gate (CONFIRMED, observed 2026-09-11): Source-code vulnerability discovery is permitted. Higher-risk requests can invoke safeguards. API fallback needs configuration; default routing is selected per model and category.. The blocked and flagged rates for Fable 5.1, Fable 5, Opus 5 and Sonnet 5 are read from the Fable 5.1 system card: Figure 3.4.2.A (page 54) gives the defensive vulnerability discovery rates, 7.0% for Fable 5.1, and Figure 3.4.3.A (page 55) the production secure coding rates, 1.03% for Fable 5.1. The card titles both charts a classifier flag rate and its caption calls the first a block rate. Anthropic's own prompting guide states the direction: "finding vulnerabilities in source code is permitted".