ModelSenseNova
SenseNova U1.5
A language and multimodal model in SenseTime's SenseNova series, with a hidden-state dimension of 4096. It can serve as a teacher model for cross-architecture semantic transfer tasks.
Timeline (last 90 days)
Sep 12
MiniMax H3 and SenseNova U1.5 are structurally completely incompatible: H3 has 264 2D weight layers and SenseNova has 593, with zero matching layers by shape, semantic type, post-transposition, or input-output dimensions.
Unverified50%