CVE-2026-116829: llama.cpp b10398 before b11301 Integer Overflow in gguf_init_from_reader
llama.cpp b10398 before b11301 contains an integer overflow vulnerability in ggufinitfromreader that allows attackers to bypass tensor element-count validation by supplying shapes whose product wraps to zero. Attackers can craft GGUF files with enormous tensor extents that overflow ggmlnbytes() and strides, causing out-of-bounds memory access when buffers are sized from wrapped byte counts.
Affected Software
Event History
Frequently Asked Questions
Which installations are affected?
ggml-org llama.cpp builds from b10398 through builds before b11301 are affected. The issue is triggered while reading GGUF files.
What does an attacker need to exploit this issue?
An attacker needs to supply a crafted GGUF file containing tensor shapes with enormous extents whose product wraps to zero. Exploitation requires the vulnerable software to process that file.
What is the practical impact of a malicious GGUF file?
The crafted tensor metadata can bypass element-count validation and cause byte-count and stride calculations to overflow. Buffers may then be sized using wrapped values, leading to out-of-bounds memory access.
What should be done if affected builds are in use?
Update to b11301 or a later build. Until updating is possible, do not process GGUF files from untrusted sources, since crafted files are the input required to trigger the vulnerable parsing path.