The vulnerability is found in the get_class_from_dynamic_module function of examples/chat_gradio_no_merge.py. A crafted value supplied in the base_model_name_or_path argument allows an attacker to inject code, which can be executed on the server. This leads directly to remote code execution, compromising confidentiality, integrity, and availability of the host system. The weakness is a combination of CWE‑74 and CWE‑94, indicating improper neutralisation of input and uncontrolled code generation.

The affected product is LlamaChinese Llama‑Chinese, with the issue present in all releases up to commit 6fa0fffb0dd82fe3cfaa1449ee54a5806d26ae9b. No official version numbers are disclosed because the project follows a rolling‑release model. Users of any snapshot containing this commit are potentially vulnerable.

The CVSS score of 5.3 indicates a moderate severity. No EPSS score is available and the vulnerability is not listed in the CISA KEV catalog, yet the exploit code has been made public and can be performed remotely. The likely attack vector is a remote request that includes a malicious base_model_name_or_path, which causes arbitrary code to be evaluated in the target environment. Because the flaw permits code injection, the risk to impacted systems is high if the vulnerable code path is exposed and no safeguards are in place.