Using the NPU on Krackan Point for LLM’s is a real gain, 2x as fast as GPU

March 16th, 2026

It took a while (about a year), but now you can use your AMD AI Krackan Point laptop for local AI on Linux. That is awesome!

There is some naming/spelling inconsistency: Krackan Point is sometimes called Kraken Point.

FastFlowLM in collaboration with AMD brought Linux support for the XDNA2 NPU (50 Tops) on AMD AI 300 series (Krackan Point).

https://fastflowlm.com/docs/install_lin/

Tried it out and yes, and the initial experience was excellent. It now performs better on Linux than on Windows for two reasons:

AMD NPU LLM benchmarks

Translating with TranslateGemma‑4B runs at roughly ~20T/s, while LLama.cpp (Vulkan) achieve about 10T/s. And Time to First Token (TTFT) is also faster.

gpt-oss:20b delivers around 19.5 T/s. (“decoding_speed_tps”:19.444631769512764}

For official benchmarks see:

https://fastflowlm.com/benchmarks/

Conclusion: AMD NPU on Linux is great for laptops

Yes, the AMD NPU is great on Linux.

Biggest win: it won’t slowdown your laptop, it runs independently of your CPU or GPU load.

This is a real advantage, together with much better power-efficiency, so it won’t drain your battery as quickly when you’re on the go.

AMD NPU on Linux for Whisper and more coming

The NPU can also accelerate speech-to-text or speech recognition (Whisper).

Image generation is in the works, together with support for newer models like Qwen3.5.

 

 

Leave a Reply