WandB

GLM 5.2, a 744B-parameter Mixture-of-Experts model with ~40B active parameters per token, was released in June 2026 under an MIT license. It uses MLA and DeepSeek Sparse Attention for a 1M context win…

Researchers have developed DiffusionBlocks, a framework that partitions transformer neural networks into independently trainable blocks to reduce memory requirements proportionally while maintaining c…