llama.cpp

a630b27d - support for GPT-OSS, Qwen 3 MoE