CVE-2025-49847
CWE-119Published: June 17, 2025· Updated: Jun 17, 2026
Official Description
llama.cpp is an inference of several LLM models in C/C++. Prior to version b5662, an attacker‐supplied GGUF model vocabulary can trigger a buffer overflow in llama.cpp’s vocabulary‐loading code. Specifically, the helper _try_copy in llama.cpp/src/vocab.cpp: llama_vocab::impl::token_to_piece() casts a very large size_t token length into an int32_t, causing the length check (if (length < (int32_t)size)) to be bypassed. As a result, memcpy is still called with that oversized size, letting a malicious model overwrite memory beyond the intended buffer. This can lead to arbitrary memory corruption and potential code execution. This issue has been patched in version b5662.
Technical Analysis
CVE-2025-49847 can be exploited remotely over the network without requiring physical or adjacent access, significantly expanding the attack surface for threat actors.
Exploitation does not require any privileges, though user interaction (Required) is needed, which slightly reduces the risk of mass automated attacks.
A successful exploit results in complete confidentiality breach (data exposure), full integrity compromise (data manipulation), availability disruption (denial of service), with a CVSS base score of 8.8.
From a weakness classification perspective (CWE-119): Buffer overflow vulnerabilities can lead to arbitrary code execution or denial of service by corrupting adjacent memory.
CVSS v3.1 Vector Breakdown
Affected Vendors & Products
Exploit & PoC Resources
Official Patches & Advisories
All References (2)
Quick Facts
Related CVEs (CWE-119)
Recommended Actions
- →Apply vendor patches immediately
- →Monitor CVE-2025-49847 in threat intel feeds
- →Review IDS/IPS signatures for exploitation attempts