---
id: CVE-2026-107183
title: >-
  llama.cpp before b11393 contains a use-after-free and double free
  vulnerability in common_chat_peg_mapper::map that allows unauthenticated
  remote attackers to corrupt heap memory via a dangling current_tool pointer
summary: >-
  llama.cpp before b11393 contains a use-after-free and double free
  vulnerability in common_chat_peg_mapper::map that allows unauthenticated
  remote attackers to corrupt heap memory via a dangling current_tool pointer.
  Attackers can submit …
severity: high
cvss: 8.1
cvssVector: 'CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H'
cwe:
  - CWE-416
vendor: ggml-org
product: llama.cpp
affected:
  - llama.cpp >= b8227 < b11393
published: '2026-10-07'
updated: '2026-10-07'
sourceUpdated: '2026-10-07T15:17:19.913'
source: NVD
sourceUrl: 'https://nvd.nist.gov/vuln/detail/CVE-2026-107183'
references:
  - url: 'https://github.com/ggml-org/llama.cpp'
    label: disclosure@vulncheck.com
  - url: >-
      https://github.com/ggml-org/llama.cpp/blob/bed0a856606ee4a24a164066f73d2379447033f5/common/chat-peg-parser.cpp#L343-L355
    label: disclosure@vulncheck.com
  - url: >-
      https://github.com/ggml-org/llama.cpp/blob/bed0a856606ee4a24a164066f73d2379447033f5/tools/server/server-schema.cpp#L321-L324
    label: disclosure@vulncheck.com
  - url: >-
      https://github.com/ggml-org/llama.cpp/commit/dbe4c3ed42343f0a7ba0fd7e808ffeaca404d29f
    label: disclosure@vulncheck.com
  - url: 'https://github.com/ggml-org/llama.cpp/pull/29942'
    label: disclosure@vulncheck.com
  - url: >-
      https://www.vulncheck.com/advisories/llama-cpp-before-b11393-use-after-free-via-common-chat-peg-mapper-chat-parser
    label: disclosure@vulncheck.com
tags:
  - nvd
  - cve.org
ssvc:
  exploitation: none
  automatable: 'no'
  technicalImpact: total
  timestamp: '2026-10-07T14:36:59.882876Z'
ingestedAt: '2026-10-07T14:33:21.952Z'
---

## Overview

llama.cpp before b11393 contains a use-after-free and double free vulnerability in common_chat_peg_mapper::map that allows unauthenticated remote attackers to corrupt heap memory via a dangling current_tool pointer. Attackers can submit a chat_parser in a POST /completion request emitting a tool-id after a tool-close tag to crash llama-server and shape a heap write primitive.

## Remediation

Refer to the linked advisories for vendor-supplied fixes and affected version ranges.