# CVE-2025-52566: llama.cpp: llama.cpp is an inference of several LLM models in C/C++

CVE-2025-52566 is fixed in llama.cpp on 1 platform by DEBIAN-CVE-2025-52566 (version 5760+dfsg-1), released June 24, 2025. It is not reported as exploited. Patch Radar priority: High.

- Web page: https://patchblox.com/security/cve/CVE-2025-52566
- Priority: High
- CVSS: 8.8 (high)
- EPSS: 0.0036
- CISA KEV: not listed
- Exploited (vendor report): no
- Publicly disclosed: no
- Published: Jun 24, 2025

## Description

llama.cpp is an inference of several LLM models in C/C++. Prior to version b5721, there is a signed vs. unsigned integer overflow in llama.cpp's tokenizer implementation (llama_vocab::tokenize) (src/llama-vocab.cpp:3036) resulting in unintended behavior in tokens copying size comparison. Allowing heap-overflowing llama.cpp inferencing engine with carefully manipulated text input during tokenization process. This issue has been patched in version b5721.

## Where it's fixed

| Platform | Product | Fixed in | Fix | Released |
|---|---|---|---|---|
| Linux | llama.cpp Debian 14 | 5760+dfsg-1 | [DEBIAN-CVE-2025-52566](https://patchblox.com/security/data/fix/linux/debian-cve-2025-52566.md) | Jun 24, 2025 |

---

Data as of 2026-11-10T01:44:33Z UTC. Source: Patchblox Patch Radar (https://patchblox.com/security/patch-radar). Patchblox finds out-of-date software on Windows, macOS and Linux and updates it through release rings: https://patchblox.com/demo

Free to use with attribution: show "Data: Patchblox Patch Radar" with a link to https://patchblox.com/security/patch-radar. Terms, including what needs a written agreement: https://patchblox.com/security/patch-radar/use-the-data

Product names, logos and trademarks shown on Patch Radar belong to their respective owners. They are used only to identify the software each security fix applies to and to help readers find their platform. Their use does not imply any affiliation with, or endorsement by, those owners, and Patchblox is not affiliated with any of them.