Benchmarks are one thing but the people really using these models, do it for a reason. Qwen team is top in open models, esp. for coding.