# llama.cpp adds new WebGPU quantization formats in b11406 release

> The open‑source llama.cpp library now supports several new low‑precision formats on WebGPU, and binaries for macOS, Linux and iOS are available.

Oossa · 2026-10-05 · https://oossa.com/en/llama-cpp-adds-new-webgpu-quantization-formats-in-b11406-release

The ggml‑org team released llama.cpp version b11406. It adds WebGPU support for the MMVQ quantization paths Q1_0, Q5_0, Q5_1, Q3_K, Q5_K, Q6_K and MXFP4. The release also provides pre‑built binaries for macOS (Apple Silicon and Intel), iOS, and several Ubuntu builds (CPU, Vulkan and CUDA).

## The facts

- Release version b11406 published on 2026-10-05
- Adds support for Q1_0, Q5_0, Q5_1, Q3_K, Q5_K, Q6_K and MXFP4 formats on WebGPU

## Why it matters

Developers can now run more memory‑efficient LLMs in browsers or on GPUs that use WebGPU, expanding where llama.cpp can be deployed.

## Sources & references

1. [ggml-org/llama.cpp b11406](https://github.com/ggml-org/llama.cpp/releases/tag/b11406) – llama.cpp, 2026-10-05

Last updated: 2026-10-05