---
id: "CVE-2025-71439"
kind: "vulnerability"
aliases: ["CVE-2025-71439"]
title: "llama.cpp: crafted GGUF model file crashes model loading via heap over-read"
severity: "medium"
cvss: 6.8
epss: null
exploited: false
status: "open"
category: "ai-tooling"
products: []
ecosystems: []
tags: ["ai-tooling", "gguf", "llama.cpp", "local-llm", "self-hosted"]
not_affected_if: ["managed-hosting", "trusted-models-only"]
published_at: "2026-10-27T01:19:03+00:00"
first_seen_at: "2026-10-27T01:34:33+00:00"
updated_at: "2026-10-27T01:36:27+00:00"
sources: [{"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2025-71439", "first_seen_at": "2026-10-27T01:34:33+00:00"}]
---

# llama.cpp: crafted GGUF model file crashes model loading via heap over-read

llama.cpp before b5649 has a heap out-of-bounds read in the vocabulary print routine, triggered by oversized special token IDs in a crafted GGUF model. An attacker who gets a team to load a malicious model crashes llama-cli or llama-server during model loading.

_Summary, check and fix written by a model from the sources; labelled as such._

## How to check

- Run `llama-server --version` or `llama-cli --version` and compare the build number against `b5649`
- Check whether models are loaded from untrusted sources, such as downloaded GGUF files from public hubs

## Fix

- Upgrade llama.cpp to build `b5649` or later
- Only load GGUF models from trusted sources until upgraded

## Timeline

- 2026-10-27T01:34:33+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2025-71439

## From the sources

llama.cpp before b5649 contains a heap out-of-bounds read vulnerability in llama_vocab::impl::print_info() that allows attackers to crash model loading via oversized special token IDs. Attackers can supply a crafted GGUF model with a short tokenizer.ggml.tokens array, causing llama-cli or llama-server to read past id_to_token and crash.

---

_Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai_