At a Glance · AI-generated

Chinese AI labs released significant model updates between September 2025 and March 2026, with DeepSeek introducing V3.2 in December featuring architectural improvements including sparse attention and reinforcement learning enhancements, while Alibaba's Qwen rolled out version 3.5 in early March. The Qwen 3.5 release generated considerable developer interest, with coverage highlighting its capabilities for local deployment and implementation guidance. Concurrent releases from other Chinese labs, including GLM-5 and MiniMax 2.5, underscored the region's competitive push in open-source large language models.