Skip to main content
CVE-2026-133710VulnerabilityPublished

llama.cpp b11003 before b12556 contains an integer overflow in the hrm_text loader load_arch_hparams() that allows

Source GCVE Database

0
Compare
Authoritative record:CVE-2026-133710· High 8.5

Not known to be exploited · CVSS 8.5 High : upgrade to b12556 soon.

Data as of 20 Nov 2026, 16:20 UTC

CVSS 4.0 base score
8.5High
How severe the flaw is on the 0–10 CVSS scale.

Thin evidence

One scorer or none
CVSS scores by source
CNA (VulnCheck)CVSS 4.08.5
CNA (VulnCheck)CVSS 3.17.8
Patch
Available – fixed in b12556
CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N

Exploitation evidence

  1. Disclosed: evidence found.

    CVE published 20 Nov 2026

  2. Detectable: no evidence on record.

    No public nuclei template indexed

  3. Weaponized: no evidence on record.

    No public exploit code indexed in Metasploit or Exploit-DB

  4. Exploited in the wild: no evidence on record.

    Not in CISA KEV and no source reports exploitation

  5. Observed by OffSeq Mirage: no evidence on record.

    No attack traffic recorded by OffSeq Mirage sensors

A missing step is not evidence of safety. Sources: cve.org (opens in a new tab), the Metasploit, Exploit-DB and nuclei-templates index as of 19 Nov 2026.

What to do

Upgrade llama.cpp to b12556.

Affected software

ProductAffectedFixed in
ggml-orgllama.cppAffected
  • ≥ b11003, < b12556
Fixed in
  • b12556

Check your own hosts for these packages with threat-finder (opens in a new tab), our free open-source scanner.

AI-suggested

If you cannot patch yet, avoid loading untrusted GGUF models or running llama.cpp versions in the affected range with untrusted input.

Description

A high-severity integer overflow vulnerability exists in llama.cpp versions from b11003 up to but not including b12556. The flaw occurs in the hrm_text loader's load_arch_hparams() function, where crafted GGUF metadata can bypass block count checks. This allows attackers to supply a GGUF model with layer factors that wrap modulo 2^32, leading to out-of-bounds writes in load_arch_tensors(), corrupting heap memory and potentially enabling code execution.

References

Community reviews

Loading reviews…