The A4B model is blazing fast and the model is super good at general inquiries. Notably worse than Qwen 3.6 for coding tasks but that says more about the Qwen model.