InternLM lmdeploy

14 published vulnerabilities name this product.

As the CNAs write it in their CVE records (vendor and product); the versions are theirs.

CVEVulnerable versionsTitleSeverityPublishedEPSSExploitation
CVE-2026-1197060.9.0 – 0.18.0LMDeploy 0.9.0 through 0.18.0 Unauthenticated RCE via /update_weights Pickle DeserializationCRITICAL 9.3
CVE-2026-33625>= 0.12.1, < 0.12.3LMDeploy vulnerable to arbitrary code execution via eval() of untrusted quant_dtype in model config loadingHIGH 8.80.4 %
CVE-2025-66455>= 0.9.2, < 0.16.0LMDeploy has Remote Code Execution by Pickle Deserialization via handle_zmq_recv in lmdeploy/lmdeploy/pytorch/disagg/conn/engine_conn.pyCRITICAL 9.81.3 %
CVE-2026-92983≤ 0.17.0InternLM LMDeploy through 0.17.0 Memory Exhaustion via Session ID MismatchHIGH 8.70.7 %
CVE-2026-92971≤ 0.17.0InternLM LMDeploy through 0.17.0 Assertion Denial of ServiceHIGH 8.70.7 %
CVE-2025-59953>= 0.9.1, < 0.10.2LMdeploy has Remote Code Execution by Pickle Deserialization via zmq_rpc.call_and_response() in InterLM/lmdeployCRITICAL 9.80.9 %
CVE-2026-768500.9.2 – < 0.16.0LMDeploy Remote Code Execution via Unsafe Pickle Deserialization in the Disaggregated Serving Peer ConnectorCRITICAL 9.31.3 %
CVE-2026-63764≤ 0.14.0LMDeploy Server-Side Request Forgery via HTTP Redirect BypassHIGH 7.70.5 %
CVE-2026-46517<= 0.12.3LMDeploy: Hardcoded trust_remote_code=True is an implicit unsafe remote-code load path with no user opt-outHIGH 7.80.2 %
CVE-2026-46432<= 0.12.3LMDeploy: Arbitrary code execution via hardcoded trust_remote_code=True in lmdeploy model initializationHIGH 7.80.1 %
CVE-2026-33626< 0.12.3LMDeploy Vulnerable to Server-Side Request Forgery (SSRF) via Vision-Language Image LoadingHIGH 7.545 %
CVE-2025-67729< 0.11.1lmdeploy vulnerable to Arbitrary Code Execution via Insecure Deserialization in torch.load()HIGH 8.80.5 %
CVE-2025-31630.7.0 ; 0.7.1InternLM LMDeploy conf.py open code injectionMEDIUM 4.80.4 %
CVE-2025-31620.7.0 ; 0.7.1InternLM LMDeploy PT File utils.py load_weight_ckpt deserializationMEDIUM 4.80.3 %