Four Chinese Labs Release Competing Open-Weight Coding Models in 12 Days — All Under 1/3 the Cost of Western Frontier

Air Street Press·medium signal

Z.ai's GLM-5.1, MiniMax M2.7, Moonshot's Kimi K2.6, and DeepSeek V4 all shipped open-weight coding models within a 12-day window, hitting roughly the same capability ceiling on agentic engineering benchmarks at dramatically lower inference cost. None costs more than a third of Claude Opus 4.7 or GPT-5.5. This coordinated wave represents a structural shift in the economics of agentic coding — the cost floor for competent code agents is collapsing. For builders: evaluate these models for cost-sensitive agent workloads; the capability-per-dollar ratio has fundamentally changed.