Skip to content
Fencer Intel
medium 5.1

CVE-2026-133612 json

SqueezeAILab SqueezeLLM CUDA Matvec Kernel quant_cuda_kernel.cu vecquant4matmul_spmv_hybrid_nuq_perchannel out-of-bounds write

published yesterday modified yesterday also EUVD-2026-131576

cvss
5.1
version 4.0, CNA
cisa kev
Not listed
not in CISA's catalog
epss
0.3%
EPSS 0.3% chance of exploitation in 30 days, percentile 25%
public exploits
0
none known

Summary

A weakness has been identified in SqueezeAILab SqueezeLLM up to a5fd71f353bf569feb7b55737c1c1493e78e8f31. This vulnerability affects the function vecquant4matmul_spmv_hybrid_nuq_perchannel of the file quant_cuda_kernel.cu of the component CUDA Matvec Kernel. Executing a manipulation of the argument full_row_indices can lead to out-of-bounds write. The attack can be launched remotely. The exploit has been made available to the public and could be used for attacks. This product does not use versioning. This is why information about affected and unaffected releases are unavailable. The project was informed of the problem early through an issue report but has not responded yet.

Is CVE-2026-133612 in your stack? Fencer shows which of your repositories are affected as soon as an advisory lands, and opens the fix as a pull request.

Check with Fencer

CISA SSVC

Exploitation
Proof of concept is public
Automatable
no
Technical impact
partial
Assessed
yesterday

Exploit prediction

0.32% chance of exploitation in the next 30 days, higher than 25% of all scored CVEs (FIRST EPSS, 2026-11-21).

Exploits and checks

  • github.com https://github.com/SqueezeAILab/SqueezeLLM/issues/77 referenced PoC, tagged "exploit" by a source, unverified

CVSS scores

assigned by version score vector
CNA 4.0 5.1 CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:P/VC:L/VI:L/VA:L/SC:N/SI:N/SA:N/E:P
CNA 3.1 5.5 CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:U/C:L/I:L/A:L/E:P/RL:X/RC:C
CNA 3.0 5.5 CVSS:3.0/AV:N/AC:L/PR:L/UI:R/S:U/C:L/I:L/A:L/E:P/RL:X/RC:C
CNA 2.0 6.5 AV:N/AC:L/Au:S/C:P/I:P/A:P/E:POC/RL:ND/RC:C

Affected products

  • SqueezeAILab SqueezeLLM

Classification

Weakness
  • CWE-787 Out-of-bounds Write
  • CWE-119 Improper Restriction of Operations within the Bounds of a Memory Buffer

References

Sources

Fencer Intel keeps one entry per vulnerability: the CVE id wins, then GHSA, and records that alias each other are merged, except that two CVEs always keep separate entries. The same data is available as JSON.