A fine-tuned version of Qwen with adaptive thinking support, improving the performance of the base model across every dimension.
