SenseTime has open-sourced its SenseNova Loop-9B-Preview model, which adopts a Looped Transformer architecture. By leveraging parameter-sharing recurrent computation modules, the model expands its effective computational depth and introduces the mHC mechanism to enhance feature interaction and training stability. It delivers performance comparable to—or even surpassing—27B models on code generation, office, and agent tasks. Ablation experiments show gains of 4.42 points on GPQA Diamond and 13.34 points on TeamBench.