DeepSeek-V3.2

About

DeepSeek-V3.2 is an advanced large language model developed by deepseek-ai, designed for high computational efficiency, superior reasoning, and agent performance. It features DeepSeek Sparse Attention (DSA) for long-context scenarios, a scalable reinforcement learning framework that enables its high-compute variant (DeepSeek-V3.2-Speciale) to surpass GPT-5 and achieve reasoning proficiency on par with Gemini-3.0-Pro, and a large-scale agentic task synthesis pipeline for integrating reasoning into tool-use. The model has demonstrated gold-medal performance in simulated International Mathematical Olympiad (IMO) and International Olympiad in Informatics (IOI) competitions. It also includes a revised chat template with tool calling and 'thinking with tools' capabilities and is released under the MIT License.