woolcoxm/LLMTest

llama.cpp backend that runs GGUF models directly on the Axera AX8850 NPU — no model conversion, no per-model compile. 24-30 t/s decode on a Raspberry Pi 5 with the CPU idle.

5 best points · 0 comments · 1 Hacker News post · 12 GitHub stars · View on GitHub

Hacker News discussions

• Axera AX8850 LLM running ggufs (5 points, 0 comments, by woolcoxm, 2026-08-26)