PT-2026-68731 · Llama.Cpp · Llama.Cpp
CVE-2026-43632
·
Published
2026-08-06
·
Updated
2026-08-08
CVSS v4.0
9.2
Critical
| Vector | AV:N/AC:H/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X |
Name of the Vulnerable Software and Affected Versions
llama.cpp versions b7492 through b9060
Description
A use-after-free issue exists in llama-server. This occurs when six tokenization endpoints—'/tokenize', '/detokenize', '/infill', '/apply-template', '/rerank', and '/anthropic/count tokens'—bypass the task queue to access
ctx server.vocab directly on HTTP worker threads. A time-of-check-time-of-use race condition allows the main thread to destroy and free vocab after the synchronization lock is released but before the handler finishes using it. This can lead to a crash or potential code execution when --sleep-idle-seconds is configured.Recommendations
Update llama.cpp to a version later than b9060.
As a temporary mitigation, avoid configuring the
--sleep-idle-seconds parameter.Fix
Use After Free
Time Of Check To Time Of Use
Found an issue in the description? Have something to add? Feel free to write us 👾
Related Identifiers
Affected Products
Llama.Cpp