HighVulnerabilityLLM-specific

CVE-2026-94626: vLLM through 0.29.0 fails to validate the tp_size parameter in kv_transfer_params on OpenAI-compatible completion…

Source: NVD(opens in a new tab)
Identifier: CVE-2026-94626
Published: September 21, 2026
Record updated: October 9, 2026

Known exploitation: Not listed in the CISA Known Exploited Vulnerabilities catalog at the last check.
EPSS: 0.6%

Summary

vLLM through 0.29.0 does not validate the tp_size parameter in kv_transfer_params on its OpenAI-compatible completion endpoints. Attackers can supply arbitrary tp_size values in prefill/decode disaggregated deployments, allocating unbounded memory and exhausting it, which triggers a kernel OOM-kill of the decode worker process.

Mitigation

The source does not state a fix yet. Check the original advisory for updates.

Details

Scores and exploitation

CVSS score: 7.5(high)
CVSS vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
Attack vector: network
Attack complexity: low
Privileges required: none
User interaction: none
EPSS: 0.6%Estimated probability that this CVE is exploited in the next 30 days, from the FIRST Exploit Prediction Scoring System. Read on November 9, 2026.
Disclosure date: September 21, 2026

Weaknesses and techniques

Weakness (CWE): CWE-789

Classification

Attack type: Denial of service
Attack sophistication: Trivial
AI component targeted: Inference
Impact: Availability
Labels: security

How this record was produced

Summaries, mitigations and labels on this page are generated automatically from the source. The linked source is authoritative.

Original source: https://nvd.nist.gov/vuln/detail/CVE-2026-94626

First tracked: September 21, 2026 at 08:10 PM ET, from NVD/CVE Database

Classifier: claude-haiku-5-5, prompt v4

Classifier confidence: 95% (self-assessed)

Topics

Inference infrastructure

Summary

Mitigation

Topics

Related items