CVE-2025-61620: vLLM resource exhaustion via malicious Jinja template
vLLM
A denial-of-service vulnerability exists in vLLM's OpenAI-compatible server when handling user-specified Jinja templates. Attackers can exhaust CPU and memory resources to cause a service outage.
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
- 6.5
- fix status
- Fix status unknown
- first report
- 7 Oct 2025
- last update
- 7 Oct 2025
- CVE ids
Timeline · 1
Sources · 1
Advisories and databases
- 2025-10-07CVE-2025-61620: vLLM: Resource-Exhaustion (DoS) through Malicious Jinja Template in OpenAI-Compatible ServerGitHub Advisory Database · github.com/advisories/GHSA-6fvq-23cw-5628
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 7 Oct 2025 · first reported 7 Oct 2025
- origin
- GitHub Advisory DatabaseVulnerability 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 07:00Z · updated 2026-10-18 07:00Z
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
# CVE-2025-61620: vLLM resource exhaustion via malicious Jinja template https://gcdradar.vercel.app/incidents/cve-2025-61620-vllm-resource-exhaustion-via-malicious-jinja · AIR-2025-0353 · closed 2025-10-07 - Lane: Patch now - Severity: 2 (Disclosed vulnerability) - Fix status: Fix status unknown - First reported: 2025-10-07 - AI role: Attack on AI - Products: vLLM - CVE ids: CVE-2025-61620 - CVSS: 6.5 ## Summary A denial-of-service vulnerability exists in vLLM's OpenAI-compatible server when handling user-specified Jinja templates. Attackers can exhaust CPU and memory resources to cause a service outage. ## What to do No action recorded; read the sources below and the vendor's advisory. ## Latest updates - 2025-10-07: CVE-2025-61620: vLLM: Resource-Exhaustion (DoS) through Malicious Jinja Template in OpenAI-Compatible Server (https://github.com/advisories/GHSA-6fvq-23cw-5628) ## Sources - 2025-10-07 GitHub Advisory Database: CVE-2025-61620: vLLM: Resource-Exhaustion (DoS) through Malicious Jinja Template in OpenAI-Compatible Server https://github.com/advisories/GHSA-6fvq-23cw-5628 Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.
Related
- closed7.5vLLM CVE-2025-59425: vLLM vulnerable to timing attack at bearer auth
Patch now · Disclosed · since 7 Oct 2025 · CVE-2025-59425
AI - closed7.1vLLM CVE-2025-6242: vLLM vulnerable to SSRF through MediaConnector
Patch now · Disclosed · since 7 Oct 2025 · CVE-2025-6242
AI - closed6.5vLLM CVE-2026-34755: vLLM Denial of Service via Unbounded Base64 Frames
Patch now · Disclosed · since 3 Apr 2026 · updated 6 Apr 2026 · CVE-2026-34755
AI - closed5.4vLLM CVE-2026-34753: vLLM Server-Side Request Forgery in batch runner
Patch now · Disclosed · since 3 Apr 2026 · updated 6 Apr 2026 · CVE-2026-34753
AI - closed6.5vLLM CVE-2026-34756: vLLM Unauthenticated Denial of Service via Parameter
Patch now · Disclosed · since 3 Apr 2026 · updated 6 Apr 2026 · CVE-2026-34756
AI - closed5.9vLLM vLLM audio downmixing differential enables moderation bypass
Patch now · Disclosed · since 17 Jul 2026 · updated 17 Jul 2026 · CVE-2026-34760
AI