High severity · published Sep 18, 2026
CVE-2026-33625
LMDeploy vulnerable to arbitrary code execution via eval() of untrusted quant_dtype in model config loading
8.8
CVSS
0.4%
EPSS
No
On KEV
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.
CVSS vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
Attack vector
Network
Complexity
Low
Privileges
None
User interaction
Required
Confidentiality
High
Integrity
High
Availability
High
Scope
Unchanged
Exploitation
Is it being used?
Exploit prediction (EPSS)0.4% · higher than 38% of CVEs
Exploited in the wildNo evidence yet
CISA KEVNot listed
Ransomware campaign useNot known
Affected products
Which versions
LM
lmdeploy · InternLM
>= 0.12.1, < 0.12.3 · source: vendor
References
NVD
https://nvd.nist.gov/vuln/detail/CVE-2026-33625
CVE.org
https://www.cve.org/CVERecord?id=CVE-2026-33625
github.com
https://github.com/InternLM/lmdeploy/releases/tag/v0.12.3
github.com
https://github.com/InternLM/lmdeploy/security/advisories/GHSA-3hmm-rh5q-gwwr
More from lmdeploy
9.8
CVE-2025-66455
LMDeploy has Remote Code Execution by Pickle Deserialization via handle_zmq_recv in lmdeploy/lmdeploy/pytorch/disagg/conn/engine_conn.py
9.8
CVE-2025-59953
LMdeploy has Remote Code Execution by Pickle Deserialization via zmq_rpc.call_and_response() in InterLM/lmdeploy
9.8
CVE-2026-76850
LMDeploy Remote Code Execution via Unsafe Pickle Deserialization in the Disaggregated Serving Peer Connector
8.8
CVE-2025-67729
lmdeploy vulnerable to Arbitrary Code Execution via Insecure Deserialization in torch.load()
8.6
CVE-2026-63764
LMDeploy Server-Side Request Forgery via HTTP Redirect Bypass