Heap Out-of-Bounds Read Vulnerability in Llama.cpp by ggml-org
CVE-2025-71439

6.8MEDIUM

Key Information:

Vendor

Ggml-org

Status
Llama.cpp
Vendor
CVE Published:
27 October 2026

What is CVE-2025-71439?

A heap out-of-bounds read vulnerability exists in Llama.cpp prior to version b5649, specifically within the llama_vocab::impl::print_info() function. This flaw allows attackers to crash the application during model loading when they supply a specially crafted GGUF model with a shortened tokenizer.ggml.tokens array. As a result, the application may read beyond the allocated memory for id_to_token, leading to potential crashes of both llama-cli and llama-server.

Affected Version(s)

llama.cpp 0

References

https://github.com/ggml-org/llama.cpp/security/advisories...
vendor-advisory
https://github.com/ggml-org/llama.cpp/pull/14145
patchissue-tracking
https://github.com/ggml-org/llama.cpp/commit/c33fe8b8c442...
patch

CVSS V4

Score:
6.8
Severity:
MEDIUM
Confidentiality:
None
Integrity:
None
Availability:
High
Attack Vector:
Local
Attack Complexity:
Low
Attack Required:
None
Privileges Required:
Undefined
User Interaction:
Unknown

Timeline

  • Vulnerability published

  • Vulnerability Reserved

Credit

Miaoqian Lin (Yuuoniy)
Jeongwoo Lee
ggerganov
.