CVE-2026-33298: llama.cpp has a Heap Buffer Overflow via Integer Overflow in GGUF Tensor Parsing
Vulnerability summary
CVE-2026-33298 is a high (CVSS 7.8) Heap-based Buffer Overflow vulnerability in ggml-org llama.cpp. EPSS estimates a 0.55% chance of exploitation in the next 30 days, higher than 47% of scored vulnerabilities. 1 public exploit is recorded.
Public exploit (GitHub, Date unknown)
Source freshness unavailable
How is this decided?Advisory lineage
Upstream 0 · Downstream 1 · Related 1Timeline
- 24-03-2026PublishedVulnerability first disclosed.
- 25-03-2026Last modifiedVulnerability information updated.
Description
llama.cpp is an inference of several LLM models in C/C++. Prior to b7824, an integer overflow vulnerability in the `ggml_nbytes` function allows an attacker to bypass memory validation by crafting a GGUF file with specific tensor dimensions. This causes `ggml_nbytes` to return a significantly smaller size than required (e.g., 4MB instead of Exabytes), leading to a heap-based buffer overflow when the application subsequently processes the tensor. This vulnerability allows potential Remote Code Execution (RCE) via memory corruption. b7824 contains a fix.
CVSS metrics
4 scores · 4 sources · 1 version- OSV · UbuntuEcosystem advisory · calculated7.8HighCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
- OSV · DebianEcosystem advisory · calculated7.8HighCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
- NVDNIST7.8HighCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
EPSS trend
Probability
0.55%
Chance of exploitation in the next 30 days
Percentile
47%
Versus all other CVEs
30-day change
Increase +0.17 pp
From 0.38% on 21-10-2026
Known exploited
Compare KEV catalogsNot in the CISA KEV catalog.
Not in the ENISA CNW catalog.
Not in the CIRCL catalog.
Public exploits
1 foundLlama.Cpp
Maturity: poc
Weaknesses
CWE · root cause- CWE-122Heap-based Buffer OverflowThe product writes past the end of a buffer that was allocated on the heap.
- CWE-190Integer Overflow or WraparoundThe product performs a calculation that can produce an integer overflow or wraparound when the logic assumes that the resulting value will always be larger than the original value. This occurs when an integer value is incremented to a value that is too large to store in the associated representation. When this occurs, the value may become a very small or negative number.
Affected systems
Known conditions: local access, no authentication, user interaction (CVSS vector). Other conditions: not stated in public data.
Debian1 product
- llama.cpp< 7965+dfsg-1
Ggml1 product
- llama.cpp< b7824
Ggml-org1 product
- llama.cpp< b7824
Ubuntu1 product
- llama.cppAll versions (No fix recorded)
4 vendors · 4 products
References
5 links- github.com/ggml-org/llama.cpp/security/advisories/GHSA-96jg-mvhq-q7q7X Refsource CONFIRMREPORTExploitVendor Advisory