VulnerabilityPublished 26 Sep 2026Updated 28 Sep 2026

CVE-2026-100651: vllm · Uncontrolled Resource Consumption

Affects vllmSource CVE Database V5

Compare

Not known to be exploited · CVSS 7.1 High : upgrade to 0.29.0 soon.

Data as of 8 Nov 2026, 23:24 UTC

CVSS 4.0 base score

7.1High

How severe the flaw is on the 0–10 CVSS scale.

EPSS: 0.3% chance of exploitation in 30 days – higher than 32% of CVEsExploit Prediction Scoring System: the likelihood attackers will use this flaw soon.
Patch: Available – fixed in 0.29.0

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

Attack Vector: Network
Attack Complexity: Low
Attack Requirements: None
Privileges Required: Low
User Interaction: None
Vulnerable system confidentiality: None
Vulnerable system integrity: None
Vulnerable system availability: High
Subsequent system confidentiality: None
Subsequent system integrity: None
Subsequent system availability: None

Bold values are the worst case: easiest to exploit or largest impact.

Attack Vector: Network: The vulnerable system is bound to the network stack and the set of possible attackers extends beyond the other options listed below, up to and including the entire Internet. Such a vulnerability is often termed “remotely exploitable” and can be thought of as an attack being exploitable at the protocol level one or more network hops away (e.g., across one or more routers).
Attack Complexity: Low: The attacker must take no measurable action to exploit the vulnerability. The attack requires no target-specific circumvention to exploit the vulnerability. An attacker can expect repeatable success against the vulnerable system.
Attack Requirements: None: The successful attack does not depend on the deployment and execution conditions of the vulnerable system. The attacker can expect to be able to reach the vulnerability and execute the exploit under all or most instances of the vulnerability.
Privileges Required: Low: The attacker requires privileges that provide basic capabilities that are typically limited to settings and resources owned by a single low-privileged user. Alternatively, an attacker with Low privileges has the ability to access only non-sensitive resources.
User Interaction: None: The vulnerable system can be exploited without interaction from any human user, other than the attacker.
Vulnerable system confidentiality: None: There is no loss of confidentiality within the Vulnerable System.
Vulnerable system integrity: None: There is no loss of integrity within the Vulnerable System.
Vulnerable system availability: High: There is a total loss of availability, resulting in the attacker being able to fully deny access to resources in the Vulnerable System; this loss is either sustained (while the attacker continues to deliver the attack) or persistent (the condition persists even after the attack has completed). Alternatively, the attacker has the ability to deny some availability, but the loss of availability presents a direct, serious consequence to the Vulnerable System (e.g., the attacker cannot disrupt existing connections, but can prevent new connections; the attacker can repeatedly exploit a vulnerability that, in each instance of a successful attack, leaks a only small amount of memory, but after repeated exploitation causes a service to become completely unavailable).
Subsequent system confidentiality: None: There is no loss of confidentiality within the Subsequent System or all confidentiality impact is constrained to the Vulnerable System.
Subsequent system integrity: None: There is no loss of integrity within the Subsequent System or all integrity impact is constrained to the Vulnerable System.
Subsequent system availability: None: There is no impact to availability within the Subsequent System or all availability impact is constrained to the Vulnerable System.

Definitions: FIRST CVSS v4.0 specification.

Exploitation evidence

Disclosed: evidence found.
CVE published 26 Sep 2026
Detectable: no evidence on record.
No public nuclei template indexed
Weaponized: no evidence on record.
No public exploit code indexed in Metasploit or Exploit-DB
Exploited in the wildExploited: no evidence on record.
Not in CISA KEV and no source reports exploitation
Observed by OffSeq MirageSeen by Mirage: no evidence on record.
No attack traffic recorded by OffSeq Mirage sensors

A missing step is not evidence of safety. Sources: cve.org (opens in a new tab), the Metasploit, Exploit-DB and nuclei-templates index as of 9 Nov 2026.

Disclosed: The CVE record is public on cve.org. Its publication is t = 0.
Detectable: A public nuclei template can find exposed instances, for defenders and attackers alike.
Weaponized: A Metasploit exploit module or Exploit-DB entry exists (Metasploit scanner modules do not count), or CISA SSVC or the CVSS exploit maturity reports a public proof of concept.
Exploited in the wild: CISA KEV lists it, CISA SSVC rates exploitation active, or a source reports attacks.
Observed by OffSeq Mirage: OffSeq Mirage honeypot sensors recorded attack traffic for this CVE, or found exposed hosts without attack traffic.

Moments are spaced evenly, not to scale; a dashed line marks a gap of six months or more. Deltas count from t = 0. Sources that give only a date (CISA KEV, Exploit-DB, or a time of exactly 00:00Z) read “same day” rather than hours. Metasploit modules carry the flaw’s disclosure date, not their release date, and nuclei templates carry no date, so both appear above but not on the timeline.

ghsa-g43f-v3gj-45x5

What to do

Upgrade vllm and vllm-project/vllm to 0.29.0.

Affected software

Product	Affected	Fixed in
vllm-projectvllm	Affected <0.29.0	Fixed in 0.29.0
GitHubvllm-project/vllm`pkg:github/vllm-project/vllm`Other threats	Affected <0.29.0	Fixed in 0.29.0

Check your own hosts for these packages with threat-finder (opens in a new tab), our free open-source scanner.

AI-suggested

Upgrade vLLM to version 0.29.0 or later, where the prompt-length validation issue is fixed. No other mitigation is indicated.

Description

vLLM before 0.29.0 fails to enforce decoder prompt-length validation on the disaggregated serving endpoint /inference/v1/generate. When the request contains a 'features' (multimodal) payload, vllm/entrypoints/serve/disagg/serving.py builds a multimodal EngineInput directly from the caller-supplied token_ids, and GenerateRequest.token_ids (vllm/entrypoints/serve/disagg/protocol.py) is not checked against model_config.max_model_len. For multimodal processors that report skip_prompt_length_check=True (for example Nemotron Parse, Whisper, and FireRedLID), InputProcessor._validate_prompt_len() returns immediately for both encoder and decoder prompts, so an overlong prompt becomes an EngineCoreRequest and reaches the worker input-batch copy into a fixed max_model_len-wide NumPy row. A client able to reach the endpoint on an affected model configuration can therefore submit an overlong token_ids list to trigger a worker failure and denial of service. Fixed in 0.29.0.

Impact

An attacker with access to the affected endpoint can cause a denial of service by submitting an overlong token_ids list, causing worker process failures. This disrupts service availability but does not indicate code execution or data compromise.

Summary

vLLM prior to version 0.29.0 does not validate the length of decoder prompts on the /inference/v1/generate endpoint when handling multimodal 'features' payloads. Specifically, for multimodal processors that skip prompt length checks (e.g., Nemotron Parse, Whisper, FireRedLID), the input length validation is bypassed, allowing an attacker to submit an excessively long token_ids list. This leads to a worker failure due to buffer overflow in a fixed-size NumPy array, resulting in denial of service. The vulnerability is addressed in vLLM 0.29.0.

References

Community reviews

Loading reviews…

Exploitation evidence

What to do

Affected software

Description

AI analysisMachine-generated, updated 3 Oct 2026, 16:54 UTC

Technical details

References

Community reviews

Related threats