DeepSeek is back at it with another tremendous model. Deepseek V3.2 is a reasoning model for agents. With DeepSeek Sparse Attention, computational complexity is reduced while performance is not compromised. It is capable of performing very well in comparison to GPT-5. DeepSeek-V3.2-Speciale comes on par with Gemini 3.0-Pro when it comes to reasoning. This model achieved Gold-medal performance in the 2025 International Mathematical Olympiad (IMO) and International Olympiad in Informatics (IOI).
🚀 Launching DeepSeek-V3.2 & DeepSeek-V3.2-Speciale — Reasoning-first models built for agents!
🔹 DeepSeek-V3.2: Official successor to V3.2-Exp. Now live on App, Web & API.
🔹 DeepSeek-V3.2-Speciale: Pushing the boundaries of reasoning capabilities. API-only for now.📄 Tech… pic.twitter.com/SC49UdmCZv
— DeepSeek (@deepseek_ai) December 1, 2025
Its chat template has a tool calling and thinking with tool. This means DeepSeek model’s thinking mode now supports tool calls, so it can improve the quality of response. You can find out more here.
