Cambricon announced Day 0 adaptation of SenseTime's newly released SenseNova Loop-9B, bringing the model's inference pipeline onto its MLU platform via vllm_mlu. The work covers static unrolling of the Loop layers, reuse of existing Qwen3-Next MLU operators, and Triton-fused optimization of mHC routing, enabling low-memory deployment of this loop-computing architecture on domestic accelerators.
SenseTime
商汤科技
Chinese AI company spanning foundation models, computer vision and computing infrastructure, developer of the SenseNova model fa
Curated Intelligence
IT之家 · 智能时代T2·Infra
58 pts
IT之家 · 智能时代T2·Models
78 pts
SenseTime released and open-sourced SenseNova Loop-9B, which reuses shared intermediate layers to expand 32 physical decoder layers into 56 layer computations, approaching 27B-class performance at roughly 9B parameters. Its official AWQ 4-bit weights are about 4.9 GB with deployment memory under 10 GB, supporting 512K context and multimodal input to sharply lower private deployment barriers.