[gitpusher42]

For Mac I would start from MLX engine. For exact model choice it is better to check bench results, and select model based on your need. A lot of good feedback about Qwen3.6, but I haven't used it in my tasks