Four Chinese Labs Just Hit Frontier-Class Coding in 12 Days
DeepSeek V4, GLM-5.1, MiniMax M2.7, and Kimi K2.6 all shipped open-weights coding models in 12 days at frontier capability — and lower inference cost than Claude or GPT.
Category
AI & ML
Models, agents, AI products, AI policy.
The AI & ML beat covers the work being shipped, not the speeches at conferences. New model releases from OpenAI, Anthropic, Google DeepMind, Meta, Mistral, and the open-source layer. Agent frameworks that actually run in production. Inference-cost trendlines and the chips underneath them. AI-native products from startups and the AI features being grafted onto incumbents.
We track the business too: where the capex is going, which AI startups are raising what at which multiples, which enterprise pilots stay pilots and which become line items. When a research paper matters for shipping software, we cover it. When it's a benchmark stunt, we say so.
Maya Iverson leads the beat. Hank Reed weighs in when the cycle gets ahead of the underlying math.