Similar to DeepSeek V3, the team released their new flagship model over a major US holiday weekend. Given DeepSeek V3.2's really good performance (on GPT-5 a...

Sebastian Raschka's Blog offers insights, tutorials, and research updates on machine learning, deep learning, and artificial intelligence. Covering topics such as neural networks, data science, and Python programming, Sebastian Raschka's Blog provides resources for students, researchers, and practitioners in the field of AI. Developers can learn about  algorithms, research methodologies, and practical applications of machine learning through Raschka's blog posts and publications.

Sebastian Raschka

DeepSeek V3.2 represents a significant evolution in open-weight language models, introducing sparse attention mechanisms (DSA) for improved efficiency and self-verification techniques from DeepSeekMath V2 for enhanced reasoning. The model maintains the Multi-Head Latent Attention (MLA) and Mixture-of-Experts architecture from V3 while adding learned token selection instead of fixed sliding windows. Training improvements include domain-specific KL tuning in GRPO, off-policy sequence masking, and hybrid RLVR combining symbolic verification with LLM-as-judge rewards. The release achieves GPT-5 level performance through architectural efficiency gains and refined reinforcement learning methods, with V3.2-Speciale offering extended thinking capabilities via reduced length penalties during training.

A Technical Tour of the DeepSeek Models from V3 to V3.2

2. Hybrid Versus Dedicated Reasoning Models

4. DeepSeek V3.2-Exp and Sparse Attention

5. DeepSeekMath V2 with Self-Verification and Self-Refinement