DeepSeek V3.2

DeepSeek V3.2 is a mixture-of-experts language model with 671 billion total parameters and 37 billion active parameters, featuring DeepSeek Sparse Attention (DSA) for efficient long-context inference and a scalable reinforcement learning framework for superior reasoning. It introduces a 'thinking with tools' capability that integrates chain-of-thought reasoning directly into tool-use scenarios, achieving GPT-5-class performance — including gold-medal results at the 2025 IMO and IOI — across advanced reasoning, coding, and agentic tasks.