...language models to run on Apple Silicon and NVIDIA GPUs without requiring Python or MLX. It supports local serving with Qwen3.5-2B-MLX-4bit and integrates with Agent Traffic Control for collaborative...