vLLM vulnerabilities

CVEs whose affected-version data names the vLLM package (pip). Each record lists the affected and patched versions; check a specific version with the dependency checker or POST /api/sbom.

82 CVEsRSS

CVE-2026-56340High· 8.8

3mo ago

vLLM versions >= 0.10.2 and < 0.13.0 are missing sparse tensor validation in multimodal embeddings processing

vLLM versions >= 0.10.2 and < 0.13.0 are missing sparse tensor validation in multimodal embeddings processing. Because PyTorch disables sparse tensor invariant checks by default, an attacker can submit crafted embedding requests with mal…

▾ Twilightvllm · vllmEPSS 0.64%via NVD

GHSA-78fp-cf4h-g36pHigh· 8.8

3mo ago

Duplicate Advisory: vLLM introduced enhanced protection for CVE-2025-62164

▾ Twilightvllm · vllmvia GHSA

GHSA-vfm7-4h43-gp6mMedium· 4.3

3mo ago

Duplicate Advisory: vLLM Vulnerable to Regular Expression Denial of Service

▾ Sunlitvllm · vllmvia GHSA

CVE-2026-54235Medium· 6.5

3mo ago

vLLM: temperature=NaN and temperature=Infinity bypass validation and propagate to GPU kernels

▾ Sunlitvllm · vllmEPSS 0.45%via OSV

CVE-2026-12491Medium· 4.8

3mo ago

vLLM: image EXIF Rotation & PNG tRNS Transparency Not Normalized, Causing Mismatch Between Model Input and Expectations

▾ Sunlitvllm · vllmEPSS 0.24%via OSV

CVE-2026-53923High· 7.5

3mo ago

vLLM: GGUF dequantize kernel int truncation exposes uninitialized GPU memory in multi-tenant serving

▾ Twilightvllm · vllmEPSS 0.48%via OSV

CVE-2026-54236Medium· 5.3PoC

3mo ago

vLLM: incomplete CVE-2026-22778 fix leaks PIL repr addresses via Anthropic router

▾ Twilightvllm · vllmEPSS 0.93%via OSV

CVE-2026-54233Medium· 6.5

3mo ago

vLLM: OOM Denial of Service via Audio Decompression Bomb

▾ Sunlitvllm · vllmEPSS 0.42%via OSV

GHSA-x8xr-mj9x-6h7wMedium· 4.8

3mo ago

Duplicate Advisory: image EXIF Rotation & PNG tRNS Transparency Not Normalized, Causing Mismatch Between Model Input and Expectations

▾ Sunlitvllm · vllmvia GHSA

CVE-2026-5497High· 7.5

3mo ago

vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method

vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method. When processing `video/jpeg` data URLs, the method …

▾ Twilightvllm · vllmEPSS 0.54%via NVD

CVE-2026-47155Medium· 6.5

3mo ago

vLLM's Artifact Pin Decay allows pinned deployments to load unpinned code, weights, and processors

▾ Sunlitvllm · vllmEPSS 0.25%via OSV

CVE-2026-9540Medium· 5.3

3mo ago

vLLM is an inference and serving engine for large language models (LLMs)

vLLM is an inference and serving engine for large language models (LLMs). From 0.1.0 to before 0.19.0, a Denial of Service vulnerability exists in the vLLM OpenAI-compatible API server. Due to the lack of an upper bound validation on the…

▾ Sunlitvllm · vllmEPSS 0.42%via NVD

CVE-2026-34755Medium· 6.5

5mo ago

vLLM is an inference and serving engine for large language models (LLMs)

vLLM is an inference and serving engine for large language models (LLMs). From 0.7.0 to before 0.19.0, the VideoMediaIO.load_base64() method at vllm/multimodal/media/video.py splits video/jpeg data URLs by comma to extract individual JPE…

▾ Sunlitvllm · vllmEPSS 0.46%via NVD

CVE-2026-34753Medium· 5.4PoC

5mo ago

vLLM: Server-Side Request Forgery (SSRF) in `download_bytes_from_url `

▾ Twilightvllm · vllmEPSS 0.25%via OSV

CVE-2026-34760High· 7.1

5mo ago

vLLM is an inference and serving engine for large language models (LLMs). From version 0.5.5 to before version 0.18.0, Librosa defaults t…

vLLM is an inference and serving engine for large language models (LLMs). From version 0.5.5 to before version 0.18.0, Librosa defaults to using numpy.mean for mono downmixing (to_mono), while the international standard ITU-R BS.775-4 sp…

▾ Twilightvllm · vllmEPSS 0.27%via OSV

CVE-2026-27893High· 8.8

5mo ago

vLLM is an inference and serving engine for large language models (LLMs)

vLLM is an inference and serving engine for large language models (LLMs). Starting in version 0.10.1 and prior to version 0.18.0, two model implementation files hardcode `trust_remote_code=True` when loading sub-components, bypassing the…

▾ Twilightvllm · vllmEPSS 1.3%via NVD

CVE-2026-25960Medium· 5.4

6mo ago

vLLM has SSRF Protection Bypass

▾ Sunlitvllm · vllmEPSS 0.54%via OSV

CVE-2026-22778Critical· 9.8PoC

7mo ago

vLLM is an inference and serving engine for large language models (LLMs)

vLLM is an inference and serving engine for large language models (LLMs). From 0.8.3 to before 0.14.1, when an invalid image is sent to vLLM's multimodal endpoint, PIL throws an error. vLLM returns this error to the client, leaking a hea…

▾ Abyssalvllm · vllmEPSS 3.8%via NVD

CVE-2026-24779High· 7.1

7mo ago

vLLM is an inference and serving engine for large language models (LLMs)

vLLM is an inference and serving engine for large language models (LLMs). Prior to version 0.14.1, a Server-Side Request Forgery (SSRF) vulnerability exists in the `MediaConnector` class within the vLLM project's multimodal feature set. …

▾ Twilightvllm · vllmEPSS 0.55%via NVD

CVE-2026-22807High· 8.8PoC

8mo ago

vLLM is an inference and serving engine for large language models (LLMs)

vLLM is an inference and serving engine for large language models (LLMs). Starting in version 0.10.1 and prior to version 0.14.0, vLLM loads Hugging Face `auto_map` dynamic modules during model resolution without gating on `trust_remote_…

▾ Midnightvllm · vllmEPSS 0.83%via NVD

CVE-2026-22773Medium· 6.5

8mo ago

vLLM is an inference and serving engine for large language models (LLMs)

vLLM is an inference and serving engine for large language models (LLMs). In versions from 0.6.4 to before 0.12.0, users can crash the vLLM engine serving multimodal models that use the Idefics3 vision model implementation by sending a s…

▾ Sunlitvllm · vllmEPSS 0.45%via NVD

CVE-2025-62372Medium· 6.5

10mo ago

vLLM vulnerable to DoS with incorrect shape of multimodal embedding inputs

▾ Sunlitvllm · vllmEPSS 0.38%via OSV

CVE-2025-62164High· 8.8

10mo ago

vLLM deserialization vulnerability leading to DoS and potential RCE

▾ Twilightvllm · vllmEPSS 0.89%via OSV

CVE-2025-62426Medium· 6.5

10mo ago

vLLM vulnerable to DoS via large Chat Completion or Tokenization requests with specially crafted `chat_template_kwargs`

▾ Sunlitvllm · vllmEPSS 0.37%via OSV

CVE-2025-61620Medium· 6.5

11mo ago

vLLM: Resource-Exhaustion (DoS) through Malicious Jinja Template in OpenAI-Compatible Server

▾ Sunlitvllm · vllmvia OSV

CVE-2025-6242High· 7.1

11mo ago

vLLM is vulnerable to Server-Side Request Forgery (SSRF) through `MediaConnector` class

▾ Twilightvllm · vllmEPSS 0.25%via OSV

CVE-2025-9141High· 8.8

1y ago

vLLM has remote code execution vulnerability in the tool call parser for Qwen3-Coder

▾ Twilightvllm · vllmvia OSV

vLLM vulnerabilities

vLLM versions >= 0.10.2 and < 0.13.0 are missing sparse tensor validation in multimodal embeddings processing

Duplicate Advisory: vLLM introduced enhanced protection for CVE-2025-62164

Duplicate Advisory: vLLM Vulnerable to Regular Expression Denial of Service

vLLM: temperature=NaN and temperature=Infinity bypass validation and propagate to GPU kernels

vLLM: image EXIF Rotation & PNG tRNS Transparency Not Normalized, Causing Mismatch Between Model Input and Expectations

vLLM: GGUF dequantize kernel int truncation exposes uninitialized GPU memory in multi-tenant serving

vLLM: incomplete CVE-2026-22778 fix leaks PIL repr addresses via Anthropic router

vLLM: OOM Denial of Service via Audio Decompression Bomb

Duplicate Advisory: image EXIF Rotation & PNG tRNS Transparency Not Normalized, Causing Mismatch Between Model Input and Expectations

vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to unbounded frame count processing in the `VideoMediaIO.load_base64()` method

vLLM's Artifact Pin Decay allows pinned deployments to load unpinned code, weights, and processors

vllm has Improper Resource Shutdown or Release

vLLM Vulnerable to Remote DoS via Special-Token Placeholders

vLLM makes Use of Uninitialized Resource

vLLM is an inference and serving engine for large language models (LLMs)

vLLM is an inference and serving engine for large language models (LLMs)

vLLM: Server-Side Request Forgery (SSRF) in `download_bytes_from_url `

vLLM is an inference and serving engine for large language models (LLMs). From version 0.5.5 to before version 0.18.0, Librosa defaults t…

vLLM is an inference and serving engine for large language models (LLMs)

vLLM has SSRF Protection Bypass

vLLM is an inference and serving engine for large language models (LLMs)

vLLM is an inference and serving engine for large language models (LLMs)

vLLM is an inference and serving engine for large language models (LLMs)

vLLM is an inference and serving engine for large language models (LLMs)

vLLM vulnerable to DoS with incorrect shape of multimodal embedding inputs

vLLM deserialization vulnerability leading to DoS and potential RCE

vLLM vulnerable to DoS via large Chat Completion or Tokenization requests with specially crafted `chat_template_kwargs`

vLLM: Resource-Exhaustion (DoS) through Malicious Jinja Template in OpenAI-Compatible Server

vLLM is vulnerable to Server-Side Request Forgery (SSRF) through `MediaConnector` class

vLLM has remote code execution vulnerability in the tool call parser for Qwen3-Coder