“The Qwen team open-sourced a 1.8 billion parameter model, which they initially considered useless for internal research, to help students and researchers with limited compute resources.”