CVE-2026-33625— LMDeploy vulnerable to arbitrary code execution via eval() of untrusted quant_dtype in model config loading

Quick assessment

Affected

InternLM lmdeploy

Exploitation

No confirmed in-the-wild exploitation; assess based on exposure

Recommended action

Check the vendor advisory and references for a fixed version. If immediate upgrade is impossible, restrict exposure and increase monitoring.

LMDeploy 是一个用于压缩、部署和服务大语言模型的工具包。版本 0.12.1 至 0.12.2 在第 620 行存在代码注入漏洞，攻击者可以通过发布一个经过特制的 HuggingFace 模型，其中包含精心构造的值，从而执行任意 Python 代码。当用户使用 LMDeploy 加载该模型时，会未经任何校验直接传入。版本 0.12.3 包含修复补丁。

CVSS 8.8 · High EPSS 0.44% · P39 Updated Sep 23, 2026

Possible ATT&CK Techniques 1 AI

T1059 · Command and Scripting Interpreter

Affected Version Matrix 1

Vendor	Product	Version Range	Status
InternLM	lmdeploy	`>= 0.12.1, < 0.12.3`	affected

Get alerts for future matching vulnerabilities Log in to subscribe

Vulnerability Information

Have questions about the vulnerability? See if Shenlong's analysis helps!

View Shenlong Deep Dive ↗

Although we use advanced large model technology, its output may still contain inaccurate or outdated information.Shenlong tries to ensure data accuracy, but please verify and judge based on the actual situation.

Vulnerability Title

LMDeploy vulnerable to arbitrary code execution via eval() of untrusted quant_dtype in model config loading

Source: CVE Program / CVE List V5

Vulnerability Description

LMDeploy is a toolkit for compressing, deploying, and serving large language models. Versions 012.1 through 0.12.2 contain a code injection vulnerability in `lmdeploy/pytorch/config.py` line 620 that allows an attacker to execute arbitrary Python code by publishing a malicious HuggingFace model with a crafted `quantization_config.quant_dtype` value. When a user loads the model with lmdeploy, the `quant_dtype` is passed to `eval(f'torch.{quant_dtype}')` without any validation. Version 0.12.3 contains a patch.

Source: CVE Program / CVE List V5

CVSS Information

CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H

Source: CVE Program / CVE List V5

Vulnerability Type

未加控制的资源消耗(资源穷尽)

Source: CVE Program / CVE List V5

Affected Products

Vendor	Product	Affected Versions	CPE	Subscribe
InternLM	lmdeploy	>= 0.12.1, < 0.12.3	-

II. Public POCs for CVE-2026-33625

#	POC Description	Source Link	Shenlong Link

AI-Generated POC Premium

No public POC found.

III. Intelligence Information for CVE-2026-33625

请登录查看更多情报信息。

Other References for CVE-2026-33625 (1)

🔗 Arbitrary code execution via eval() of untrusted quant_dtype in model config loading · Advisory · InternLM/lmdeploy · GitHub Read full article x_refsource_CONFIRM

https://www.cve.org/CVERecord?id=CVE-2026-33625

IV. Related Vulnerabilities

Same product: lmdeploy

Same vendor: InternLM

Same weakness: CWE-400