CVE-2026-43631
llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in the vocab pointer of llama-server when the --sleep-idle-seconds feature is enabled, allowing unauthenticated remote attackers to execute arbitrary code. Attackers can trigger the vulnerability by sending requests to affected endpoints while the server transitions to sleep mode, causing concurrent worker threads to dereference a freed vocab pointer that can be reclaimed with attacker-controlled data to achieve remote code execution.
Attack profile
The conditions required to exploit this vulnerability and its potential impact.
Affected products
Product applicability statements supplied with the NVD record.
Weakness classification
CWE categories help security teams group the underlying software weakness.
Source references
External advisories, patches and technical reports attached to this CVE record.