Skip to main content
← Back to sources

DeepSeek Releases V3.2 and V3.2-Speciale Models Rivaling Frontier Competitors

Published 2025-12-01Foundation ModelsHigh

Summary

Chinese AI lab DeepSeek released two new models on December 1, 2025: DeepSeek-V3.2 and DeepSeek-V3.2-Speciale. DeepSeek-V3.2 is the official successor to V3.2-Exp, balancing inference speed and reasoning capability across web, mobile, and API deployments. The higher-end V3.2-Speciale pushes the boundaries of reasoning performance, rivaling Google's Gemini-3.0-Pro in capability, though with higher token usage and initially API-only access. DeepSeek-V3.2 introduces a new mechanism called DeepSeek

Alignment: New signal not yet covered
Related Positions: ai-governance-and-risk.md
deepseekfoundation-modelschinamodel-releasebenchmarkreasoninggeopoliticsopen-sourcesparse-attention
DeepSeek Releases V3.2 and V3.2-Speciale Models Rivaling Frontier Competitors — Intelligence — Agentic Developer Tools Radar · Signal