# vLLM denial of service via prompt embeds on M-RoPE models

Advisory record · `GHSA-33cg-gxv8-3p8g`

- **Identifiers:** `GHSA-33cg-gxv8-3p8g` · `CVE-2026-55514` · `PYSEC-2026-2303`
- **State:** Confirmed
- **Severity:** CVSS 7.1 High
- **Critical:** No
- **Weakness (CWE):** [CWE-617](https://cwe.mitre.org/data/definitions/617.html)
- **Tags:** [Denial of service](https://agentprime.io/advisories/tag/dos.md)
- **Published:** 20 Jul 2026
- **Updated:** 10 Sept 2026
- **Reviewed by a person:** No

Published by the GitHub Advisory Database at 20 Jul 2026, 19:13 UTC. agentprime has not analysed this yet, so whether it affects any app is not known.

## Affected packages

- `vllm` (PyPI): affected `>=0.12.0, <0.24.0`; fixed in `0.24.0`

## Machine-written summary

Summary written by Claude Haiku from the source advisory. Not reviewed by a person.

> Versions of vLLM from 0.12.0 before 0.24.0 are affected when prompt embeds are enabled on a model that uses M-RoPE. A remote user who can reach the /v1/completions endpoint can send a request with prompt embeds and no prompt text, which triggers a failed assertion that crashes the engine and shuts down the server (denial of service). Upgrade to vLLM 0.24.0 or later. Disabling prompt embeds on affected M-RoPE models is a possible workaround.

Written 4 Nov 2026.

## Not analysed yet

We haven’t published our analysis of this Advisory yet. Once a person has reviewed it, this page will say what’s vulnerable, what has to be true for your app to be affected, and how to check.

## Severity and exploitation

- 7.1 High, CVSS 4.0: `CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N`, from CNA (GitHub_M)

- **CISA KEV:** Not listed.
- **EPSS:** 0.0066, the 52nd percentile, on 15 Nov 2026: FIRST’s estimate of the chance of exploitation in the next 30 days.

## Timeline

- 31 Oct 2026, 14:29 UTC: First seen by agentprime
- 31 Oct 2026, 14:29 UTC: [Confirmed by OSV](https://osv.dev/vulnerability/GHSA-33cg-gxv8-3p8g)

## Sources

- [GitHub Advisory Database: GHSA-33cg-gxv8-3p8g](https://github.com/advisories/GHSA-33cg-gxv8-3p8g)
- [OSV: GHSA-33cg-gxv8-3p8g](https://osv.dev/vulnerability/GHSA-33cg-gxv8-3p8g)
- <https://github.com/vllm-project/vllm/security/advisories/GHSA-33cg-gxv8-3p8g>
- <https://nvd.nist.gov/vuln/detail/CVE-2026-55514>
- <https://github.com/vllm-project/vllm/pull/45252>
- <https://github.com/vllm-project/vllm/commit/470229c37efaf69c86e8bc97482b0b1ff7551c65>
- <https://github.com/pypa/advisory-database/tree/main/vulns/vllm/PYSEC-2026-2303.yaml>
- <https://github.com/vllm-project/vllm>
- <https://github.com/vllm-project/vllm/releases/tag/v0.24.0>

Facts on this page come from the GitHub Advisory Database (CC-BY 4.0), OSV and the other sources listed, as published. We don’t reproduce full advisory descriptions, which can contain exploit code.

## Check whether this affects you

A Match means an installed version is in the affected range above. Most don’t affect the app; agentprime hasn’t analysed this one yet.

- With the agentprime MCP server (<https://api.agentprime.io/mcp>), ask your agent to check the project: `check_lockfile` sends a lockfile, `check_packages` a list of installed packages, and both return the Matches and the upgrades that clear them. `get_advisory` returns this record. Setup: <https://agentprime.io/mcp.md>
- In a browser: <https://agentprime.io/check/> reads your lockfile in the browser, never uploading it.
- Whether it touches your code, with file and line: the Claude Code plugin, <https://agentprime.io/plugin.md>
- As JSON: <https://api.agentprime.io/v1/advisories/GHSA-33cg-gxv8-3p8g>, or `POST https://api.agentprime.io/v1/match` with your packages (<https://agentprime.io/api.md>)

---

The web page: <https://agentprime.io/advisories/GHSA-33cg-gxv8-3p8g/>. Everything for agents: <https://agentprime.io/llms.txt>.