CVE-2026-43631: llama.cpp b7492–b9060 Use-After-Free RCE via llama-server
llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in the vocab pointer of llama-server when the --sleep-idle-seconds feature is enabled, allowing unauthenticated remote attackers to execute arbitrary code. Attackers can trigger the vulnerability by sending requests to affected endpoints while the server transitions to sleep mode, causing concurrent worker threads to dereference a freed vocab pointer that can be reclaimed with attacker-controlled data to achieve remote code execution.
Security readout for executives and security teams
Plain-English summary
Affected llama-server deployments may let an unauthenticated network attacker take control of the server. The flaw occurs during sleep-mode transitions and can corrupt memory when concurrent workers use released vocabulary data. Compromise could expose model data, alter services, or disrupt availability.
Executive priority
Treat internet-accessible affected servers as an immediate remediation priority. Quickly disable the triggering feature or isolate the service, then establish an authoritative upgrade path. Internal-only deployments still warrant prompt action because successful exploitation could provide full control of the serving process.
Technical view
CVE-2026-43631 is a race condition and use-after-free involving llama-server’s vocab pointer when --sleep-idle-seconds is enabled. Requests arriving during a sleep transition may cause worker threads to dereference freed memory. The supplied description states that attacker-controlled reclamation can lead to remote code execution in builds b7492 through b9060.
Likely exposure
Highest exposure is in network-accessible llama-server instances running builds b7492 through b9060 with --sleep-idle-seconds enabled. Authentication or network controls may reduce reachability, but the described vulnerability itself requires no privileges or user interaction. Instances without that feature enabled are not described as vulnerable.
Exploitation context
The CVSS 4.0 score is 9.2. Exploitation requires network access, a sleep transition, and favorable concurrency or timing conditions. The bundle does not report CISA KEV inclusion or provide evidence of active exploitation. High attack complexity should not be treated as protection for exposed services.
Researcher notes
The reported weakness combines CWE-362 and CWE-416: concurrent activity during sleep transition leaves workers using a freed vocab pointer. CVSS reflects attack requirements through AC:H and AT:P. The bundle’s structured affected entry lists only b7492, while its title and description state b7492 through b9060; confirm exact version boundaries with authoritative guidance.
Mitigation direction
Disable --sleep-idle-seconds on affected servers as an immediate exposure-reduction measure.
Restrict llama-server access to trusted networks and authenticated clients.
Review the cited security-patch repository and vendor guidance before applying a permanent fix.
Upgrade only to a release explicitly confirmed by authoritative guidance as not affected.
Restart affected services after configuration or binary changes to clear vulnerable process state.
Validation and detection
Inventory llama.cpp build identifiers across every llama-server deployment.
Confirm whether --sleep-idle-seconds is present in startup configuration or service definitions.
Identify instances reachable from untrusted networks, including through proxies or gateways.
Review logs for crashes or abnormal behavior around sleep transitions and concurrent requests.
Verify mitigations in staging, then confirm the deployed build and runtime arguments.
Generated from the cited source records. This long-tail analysis has not been individually reviewed by a named human.
Potential ATT&CK relevance
Conservative CVE-to-ATT&CK context
These mappings and lookup hints may be relevant to the vulnerability behavior, CWE, affected product, or exposure path. Glexia-inferred context is not an official MITRE, ATT&CK, CWE, or CVE Program mapping.
ATT&CK lookup starting points
Use these exact CWE pages and searches to review the Glexia ATT&CK library from this CVE's weakness and description context.
cwe · low confidence lookup
CWE-362: Exact CWE lookup
Use the exact CWE identifier as the starting point before reviewing related ATT&CK behavior. Open the exact CWE lookup page first, then review the ATT&CK searches from that MITRE weakness context. This is a Glexia lookup hint, not an official ATT&CK mapping.
Use the exact CWE identifier as the starting point before reviewing related ATT&CK behavior. Open the exact CWE lookup page first, then review the ATT&CK searches from that MITRE weakness context. This is a Glexia lookup hint, not an official ATT&CK mapping.
The CVE wording references code or command execution, so execution technique review may help defensive triage. This is a Glexia inferred lookup path, not an official MITRE, ATT&CK, or CVE Program mapping.
These fields come from the CVE record and ADP containers, not from Glexia's Take. They preserve time-varying source decisions such as CISA SSVC, KEV status, CVSS metrics, and provider references.
2CVSS vectors
3Timeline events
1ADP providers
3Source links
SSVC decision data
CISA-ADPCISA Coordinator
Timestamp
Version
2.0.3
Exploitation: noneAutomatable: noTechnical Impact: total
CVSS vector scores
2 official scores
We collect every scored CVSS vector available in the official CNA and ADP containers. When more than one version is present, the table keeps the source vectors side by side instead of collapsing them into the highest score.
CWE links open Glexia weakness intelligence pages with official CWE context, developer remediation guidance, and related CVE mappings.
CWE-362 · source CWE mapping
Concurrent Execution using Shared Resource with Improper Synchronization ('Race Condition')
Concurrent Execution using Shared Resource with Improper Synchronization ('Race Condition') represents a recurring weakness pattern that can create exploitable paths when design, validation, or implementation controls are missing.
Use After Free represents a recurring weakness pattern that can create exploitable paths when design, validation, or implementation controls are missing.