# llama.cpp speeds up quantized prefill on Intel Arc GPUs

> Release b11979 adds SYCL dequantization changes that improved prefill speeds in tests on an Intel Arc Pro B70, while decode stayed unchanged.

Oossa · 2026-10-29 · https://oossa.com/en/llama-cpp-speeds-up-quantized-prefill-on-intel-arc-gpus

The ggml-org/llama.cpp b11979 release changes how its SYCL kernels convert several quantized model formats on Intel GPUs. On an Arc Pro B70, the project measured faster prefill—the processing of prompt text—when tensors used a reordered layout, with gains varying by format and model.

The largest reported gain on that path was 28.7% for a Q6_K 27-billion-parameter model. For MXFP4 mixture-of-experts models, prefill improved 57% to 61% in the reported test, including with a fresh layout; decode speed was unchanged. The measurements are project-reported, and perplexity results were unchanged or within run-to-run noise.

## The facts

- llama.cpp b11979 reports a 28.7% prefill gain for a Q6_K 27B model on an Arc Pro B70.
- MXFP4 MoE prefill improved 57% at pp2048 and 61% at pp512; decode was unchanged.

## Why it matters

People running these model formats on an Intel Arc Pro B70 may see shorter prompt-processing time, though the gains depend on the model and layout and do not speed up decoding.

## Sources & references

1. [ggml-org/llama.cpp b11979](https://github.com/ggml-org/llama.cpp/releases/tag/b11979) – llama.cpp, 2026-10-29

Last updated: 2026-10-29