The llama.cpp project has released version b12001, which includes support for the RDNA4_M family of GPUs. This update specifically adds definitions for gfx1170 and gfx1171, aligning them with the RDNA4 architecture. The release was assisted by GitHub Copilot and Codex, and co-authored by Mario Limonciello from AMD. AI
IMPACT Enhances hardware compatibility for local LLM inference, potentially improving performance on newer AMD GPUs.
RANK_REASON Update to an open-source project for running LLMs, adding hardware support.
Read on llama.cpp — Releases →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →