vLLM Mooncake connector GPU memory exhaustion vulnerability
vLLM
The vLLM Mooncake connector fails to properly manage GPU KV cache block ownership when concurrent child requests share transfer IDs. Attackers can cause GPU memory exhaustion by accumulating orphaned cache blocks until process restart.
disclosed CVENo confirmed exploitation, so this record is tracked as a disclosed vulnerability and not counted as an incident. It moves into the counts if CISA or the vendor confirms exploitation. Why
- CVSS
- 8.7
- EPSS
- 0.01 · top 50% · 2026-11-13
- fix status
- Fix status unknown
- first report
- 21 Sep 2026
- last update
- 21 Sep 2026
- CVE ids
Timeline · 1
Sources · 1
Advisories and databases
- 2026-09-21CVE-2026-94627: vLLM Mooncake connector through 0.29.0 fails to properly manage GPU KV cache block ownership when concurrent child requests share…NVD · nvd.nist.gov/vuln/detail/CVE-2026-94627
Titles and links belong to their publishers. The summary above is written by gcdRadar in its own words; data is CC BY-SA 4.0.
Classification
- lane
- Patch now
- AI role
- Attack on AI
- categories
- Infrastructure vulnerability
- severity
- 2 · Disclosed vulnerability
- fix status
- Fix status unknown
- targets
- Framework / tooling
- owasp llm top 10
- LLM03:2025 Supply Chain
- mitre atlas
- no technique for these categories
- dates
- happened 21 Sep 2026 · first reported 21 Sep 2026
- origin
- NVDVulnerability record from NVD / the GitHub Advisory Database. This product uses the NVD API but is not endorsed or certified by the NVD.
- record
- confidence 0.95 · created 2026-10-18 06:33Z · updated 2026-10-18 06:33Z
Lane, AI role, categories and severity come from the classifier; the OWASP and ATLAS rows are derived from the categories. How these are assigned
Copy a brief
This record as a short Markdown brief, for a ticket, a chat or a coding agent.
the brief as text
# vLLM Mooncake connector GPU memory exhaustion vulnerability https://gcdradar.vercel.app/incidents/vllm-mooncake-connector-gpu-memory-exhaustion-vulnerability · AIR-2026-0156 · closed 2026-09-21 - Lane: Patch now - Severity: 2 (Disclosed vulnerability) - Fix status: Fix status unknown - First reported: 2026-09-21 - AI role: Attack on AI - Products: vLLM - CVE ids: CVE-2026-94627 - CVSS: 8.7 ## Summary The vLLM Mooncake connector fails to properly manage GPU KV cache block ownership when concurrent child requests share transfer IDs. Attackers can cause GPU memory exhaustion by accumulating orphaned cache blocks until process restart. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2026-09-21: CVE-2026-94627: vLLM Mooncake connector through 0.29.0 fails to properly manage GPU KV cache block ownership when concurrent child requests share… (https://nvd.nist.gov/vuln/detail/CVE-2026-94627) ## Sources - 2026-09-21 NVD: CVE-2026-94627: vLLM Mooncake connector through 0.29.0 fails to properly manage GPU KV cache block ownership when concurrent child requests share… https://nvd.nist.gov/vuln/detail/CVE-2026-94627 Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- closedvLLM vLLM Denial of Service through Multimodal UUID Cache Desynchronization
Patch now · Disclosed · since 21 Oct 2026
AI - closedvLLM vLLM Denial of Service via Unvalidated mm_processor_kwargs
Patch now · Disclosed · since 21 Oct 2026
AI - closed6.5vLLM CVE-2026-114050: vLLM multimodal UUID cache identity flaw leads to engine crash
Patch now · Disclosed · since 21 Oct 2026 · updated 23 Oct 2026 · CVE-2026-114050
AI - closed7.1vLLM CVE-2026-114054: vLLM cache aliasing vulnerability exposes processed media and KV state
Patch now · Disclosed · since 21 Oct 2026 · CVE-2026-114054
AI - closed7.5vLLM CVE-2026-114051: vLLM input validation check omission causes denial of service
Patch now · Disclosed · since 21 Oct 2026 · CVE-2026-114051
AI - closed6.3vLLM CVE-2026-114049: vLLM cross-request mask leak vulnerability
Patch now · Disclosed · since 21 Oct 2026 · CVE-2026-114049
AI