The latest llama.cpp release introduces support for the Minimax2 Eagle3 model, enhancing its inference capabilities for new AI architectures. This update also expands compatibility across various platforms, including macOS Apple Silicon with KleidiAI, Ubuntu with Vulkan and ROCm 7.2, and Windows with CUDA 12/13 and HIP.
llama.cpp continues to broaden its model and hardware support, enabling wider deployment of new AI models across diverse compute environments.