Open Models have crossed a threshold

Recent evaluations show that open weight large language models like GLM-5 and MiniMax M2.7 perform comparably to closed frontier models on key agent tasks, offering lower cost and latency advantages for production use.