CVE-2026-73557: vLLM is an inference and serving engine for large language models

Routine

MEDIUM 5.3REMOTEUPGRADE

CVE-2026-73557 is a medium-severity flaw (CVSS 5.3) affecting vllm. Below: why it ranks where it does, which versions are affected, what to do, and the exact commands.

vLLM is an inference and serving engine for large language models. From 0.20.2rc0 until 0.26.0, safe_load_prompt_embeds in vllm/renderers/embed_utils.py uses torch.sparse.check_sparse_tensor_invariants, whose process-global save, enable, and restore state can be raced by concurrent prompt_embeds parts submitted to POST /v1/chat/completions through AsyncMultiModalItemTracker.resolve_items, asyncio.gather, and the default executor, allowing an inva

Why this priority

Remotely exploitable, no login, no user action

A fixed release or distro patch exists.

Risk facts

CVSS score	5.3 (Medium)
Attack vector	Network (remote)
Attack complexity	Low
Privileges needed	None
User interaction	None
EPSS exploit chance	0.2% (top 84%)
CISA exploited list	No
Weakness	CWE-362
Published	2026-08-13
Last modified	2026-10-15

Publisher (CNA): GitHub, Inc. · Vendor: vllm

Last updated 2026-10-15 · Sources: CVE Program, NVD, CISA KEV, EPSS and vendor advisories · How we score

Am I affected?

Timeline

2026-08-13 15:20 UTCPublished CVE record published
2026-10-15 20:42 UTCUpdated Record last modified

vllm vllm UPGRADE

Affected: >= 0.21.0 and < 0.26.0

Fixed in: 0.26.0

Safe to downgrade to: a release before 0.21.0

1. Check installed version

apt-cache policy vllm || rpm -q vllm || vllm --version

2. UPGRADE to 0.26.0 (Debian / Ubuntu)

sudo apt update && sudo apt install --only-upgrade vllm

2. UPGRADE (RHEL / Amazon Linux / Fedora)

sudo dnf upgrade vllm

Note

Distros often backport fixes without changing the version number. Confirm with the distro advisory or changelog.

3. If no update is available yet, DOWNGRADE to a release before 0.21.0 (Debian / Ubuntu)

apt-cache madison vllm sudo apt install vllm=<VERSION_BEFORE_0.21.0> sudo apt-mark hold vllm

3. If no update is available yet, DOWNGRADE (RHEL / Amazon Linux / Fedora)

dnf --showduplicates list vllm sudo dnf downgrade vllm-<VERSION_BEFORE_0.21.0> sudo dnf versionlock add vllm

Downgrade risk

Downgrading can break dependent software and may reopen older bugs. Read the distro advisory first and test on a non-production host.

After downgrade

Restart the service, and keep the package held until a fixed release ships.

Verify at the official source NVD · CVE Program · Report a problem with this page
Scores and fixes here are computed automatically from those and other public sources. Confirm with the vendor advisory before changing production systems.

Quick answers

How do I fix CVE-2026-73557?

vllm: upgrade to 0.26.0. Exact commands for each OS are on this page.

Is CVE-2026-73557 being exploited?

CVE-2026-73557 is not on the CISA Known Exploited Vulnerabilities list at the time of the last update. Check again, as this can change.

What does CVE-2026-73557 affect?

CVE-2026-73557 affects vllm. Severity is medium (CVSS 5.3). vLLM is an inference and serving engine for large language models. From 0.20.2rc0 until 0.26.0, safe_load_prompt_embeds in vllm/renderers/embed_utils.py uses torch.sparse.check_sparse_tensor_invariants, whose process-global save, enable, an

How urgent is CVE-2026-73557?

Priority 28/100. Remotely exploitable, no login, no user action

Related CVEs

CVE-2026-5497 HIGH 7.5 UPGRADE vLLM versions 0.8.0 and later are vulnerable to an Out-of-Memory (OOM) Denial of Service (DoS) attack due to u
CVE-2026-48746 CRITICAL 9.1 UPGRADE vLLM is an inference and serving engine for large language models (LLMs)
CVE-2026-54232 HIGH 8.8 UPGRADE vLLM is an inference and serving engine for large language models (LLMs)
CVE-2026-41523 HIGH 7.5 UPGRADE vLLM is an inference and serving engine for large language models (LLMs)
CVE-2026-54236 MEDIUM 5.3 UPGRADE vLLM is an inference and serving engine for large language models (LLMs)

References

← Back to latest CVEs