sensenova/SenseNova-Loop-9B: Open-source 9B looped multimodal model release
SenseNova has released SenseNova-Loop-9B, a 9-parameter multimodal model that uses a looped architecture and manifold-constrained hyper-connections to expand computational depth within a compact parameter budget. The model supports text and images, handles contexts up to 512K tokens, and is competitive with larger models on various coding and data analysis benchmarks. Model weights, code, and deployment examples are available to the community. Open article →