CVE-2026-110695: Buffer Overflow in turboderp ExLlamaV2

This vulnerability exists in turboderp ExLlamaV2 versions 0.3.0, 0.3.1, and 0.3.2 within the STFile.get_tensor function of exllamav2/stloader.py. Manipulating the data_offsets argument can trigger a buffer overflow condition. The flaw is remotely exploitable without privileges or user interaction, with low complexity and partial impact on confidentiality, integrity, and availability. The issue was reported early to the project but remains unaddressed, and public exploit code is available.

Successful exploitation can lead to a buffer overflow, potentially allowing an attacker to execute arbitrary code or cause a denial of service remotely. The CVSS 4.0 score of 5.3 reflects medium severity with network attack vector, no privileges required, and partial impacts on confidentiality, integrity, and availability. No in-the-wild exploitation is currently known.

No official fix or patch is currently available, and the vendor has not responded to the issue report. Users should monitor the vendor's communications for updates. Until a patch is released, avoid exposing vulnerable versions (0.3.0, 0.3.1, 0.3.2) to untrusted networks or inputs that could manipulate the data_offsets argument.