SenseNova-U1: Unifying Multimodal Understanding and Generation with NEO-Unify Architecture
SenseNova released SenseNova-U1 with 4 MoT multimodal models. The post lists 8B and A3B variants with GitHub and HuggingFace weight links. The key claim is a monolithic architecture instead of adapters; benchmarks are not disclosed.
Why it matters: SenseNova-U1 handles text and images inside the model with no external image adapter, and releases 8B and A3B MoT weights, with A3B smaller and suited to local runs.