§ feed · storyline
From DeepSeek V3 to V3.2: Architecture, Sparse Attention, and RL Updates
DeepSeek updates its V3 open-weight model to V3.2 with changes to architecture, sparse attention mechanisms, and reinforcement learning training.
Understanding How DeepSeek's Flagship Open-Weight Models Evolved
§ sources1 publication · timeline below
- magazine.sebastianraschka.comFrom DeepSeek V3 to V3.2: Architecture, Sparse Attention, and RL Updatesprimary