CVE-2026-94623

Summary

The vLLM library has a denial of service vulnerability in its NIXL connector's prefix caching implementation that fails to properly validate block counts when handling multiple prompts in prefill/decode deployments, leading to assertion failures and worker termination.

Exploitability

Exploitation is relatively straightforward as attackers only need to submit completion requests with multiple prompts of varying lengths to trigger the assertion failure in the NixlBaseConnectorWorker._apply_prefix_caching method.

Blast radius

If exploited, this vulnerability could cause decode workers to terminate and become unavailable until restarted, potentially disrupting service availability for all users relying on the affected vLLM deployment.

Detection

No reliable host or network indicator is derivable from the published description.

Prioritized remediation

Upgrade to vLLM version 0.29.1 or later to address this denial of service vulnerability in the NIXL connector's prefix caching implementation.

Analysis generated locally by garnet-coder (no data left the box). AI-assisted — verify against primary sources before acting.