A frontier model refuses every task on a cybersecurity benchmark

Claim

On a private cybersecurity benchmark, Kimi K3 emerged as the workhorse for hunting vulnerabilities at near-frontier recall for a fraction of the cost, GPT-5.6 Sol took the crown at seven times the price, GLM 5.2 undercut everyone, and Fable 5 refused 100% of the tasks, turning safety into a service outage.

Reported metric: 100% refusal

Reporting date: 2026-07-19 (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

• Welcome to July 19, 2026 (theinnermostloop.substack.com)