CVE-2026-116502

MediumCVSS 4.4

Published: 26/10/2026(UTC)Updated: 02/11/2026(UTC)Translated: 27/10/2026(UTC)NVD NIST

Exploitation Probability (EPSS)

Low risk

0.15%

4th percentile - higher than 4% of all known CVEs

Summary

In ExLlamaV2 up to 0.3.2, the function QMatrix::make_sequential in exllamav2/exllamav2_ext/cuda/q_matrix.cu (GPTQ Index Handler component) can cause an out-of-bounds write when the g_idx argument is manipulated. The attack requires local access. The project was informed but has not responded.

Risk Assessment

A local attacker can exploit the flaw to corrupt memory, potentially leading to a crash or code execution in the process context.

Recommendation

Restrict access to the ExLlamaV2 runtime environment and avoid processing untrusted input. Monitor the project for a fix.

Other vulnerabilities in ExLlamaV2

CVE-2026-110695Medium
A vulnerability was detected in turboderp ExLlamaV2 up to 0.3.2. Affected is the function STFile.get_tensor of the file exllamav2/stloader.py of the component Safetensors Tensor Loader. Performing a manipulation of the argument data_offsets results in buffer overflow. The attack is possible to be carried out remotely. The exploit is now public and may be used.

Original NVD description (English source)

A flaw has been found in turboderp-org ExLlamaV2 up to 0.3.2. The impacted element is the function QMatrix::make_sequential of the file exllamav2/exllamav2_ext/cuda/q_matrix.cu of the component GPTQ Index Handler. Executing a manipulation of the argument g_idx can lead to out-of-bounds write. The attack needs to be launched locally. The project was informed of the problem early through an issue report but has not responded yet.

Vulnerability data from NVD (NIST) · CISA KEV · EPSS