Three Chinese AI labs, MiniMax, DeepSeek, and Alibaba, are releasing new programming-focused large language models that rival top Western providers. MiniMax's M3, launched June 1, 2026, offers a 1 million token context window, multimodal capabilities, and plans for open-sourcing weights. DeepSeek's V4-Pro excels in reasoning for complex coding tasks and has a history of open-sourcing weights, while Alibaba's Qwen 3.7 also features a 1 million token context window but currently has closed weights. These models provide more accessible and flexible options for developers building coding agents, especially for those seeking lower costs or reduced vendor dependency. AI
IMPACT Accelerates development of coding agents by providing powerful, accessible, and open-weight alternatives to proprietary models.
RANK_REASON New model releases from multiple labs with significant capabilities and open-weight plans. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →