llama.cpp

ab69c58a - support for GPT-OSS, Qwen 3 MoE