MEDIUM 6.5

PYSEC-2026-2303

Details

vLLM is a library for LLM inference and serving. From 0.12.0 to before 0.24.0, sending a pure prompt embeds payload in a /v1/completions request with a model using M-RoPE causes EngineCore to fail an assertion and fatally crash, shutting down the entire server application. Any remote user who is authorized to make a /v1/completions request can make such a request and induce a crash. This issue is fixed in version 0.24.0.

Are you affected?

Enter the version of the package you're using.

Affected packages

PyPI / vllm

Introduced in: 0.12.0 Fixed in: 0.24.0

Fix pip install --upgrade 'vllm>=0.24.0'

References

https://github.com/vllm-project/vllm/releases/tag/v0.24.0 [ADVISORY]
https://github.com/vllm-project/vllm/security/advisories/GHSA-33cg-gxv8-3p8g [ADVISORY]
https://github.com/vllm-project/vllm/commit/470229c37efaf69c86e8bc97482b0b1ff7551c65 [FIX]
https://github.com/vllm-project/vllm/pull/45252 [FIX]