# llama.cpp adds Hexagon SSM_SCAN operation support

> The open‑source llama.cpp library now supports the SSM_SCAN op on Qualcomm Hexagon NPU, improving performance on Snapdragon devices.

Oossa · 2026-11-13 · https://oossa.com/en/llama-cpp-adds-hexagon-ssm-scan-operation-support

The llama.cpp project released a new version that adds support for the SSM_SCAN operation on Qualcomm's Hexagon NPU. The update includes kernel optimizations, DMA pipeline changes, and HMX support. It also expands platform builds, adding CUDA 13 and new Snapdragon CPU‑GPU‑NPU configurations. The change was co‑authored by Max Krasnyansky of Qualcomm.

## The facts

- Adds Hexagon SSM_SCAN op support (PR #31921)
- Supports CUDA 13.5 libraries on Ubuntu and Windows

## Why it matters

Developers can run llama.cpp models faster on Snapdragon phones that have Hexagon NPUs.

## Sources & references

1. [ggml-org/llama.cpp b12373](https://github.com/ggml-org/llama.cpp/releases/tag/b12373) – llama.cpp, 2026-11-13

Last updated: 2026-11-13