A recent Mozilla report reveals that the performance gap between US-based closed frontier AI models and Chinese open-weights models has narrowed to approximately four months. Technical benchmarks indicate that while closed models like Anthropic’s Fable 5 maintain a lead in complex tasks, open models are rapidly closing the distance. For example, Moonshot AI’s Kimi K3 achieves a composite score on the Artificial Analysis Intelligence Index only three points behind Fable 5, while costing only 30% as much. When evaluated using a neutral harness via Vals AI's Terminal-Bench 2.1, Z.ai’s GLM 5.2 performed within one point of Anthropic’s Claude Opus 4.7 and 4.8, despite costing five times less per task. This gap is most visible in task duration; currently, the best closed models can reliably handle tasks 1.7 times longer than the best open models—roughly a 12-hour task versus a 7-hour task. Consequently, organizations are increasingly adopting open models for routine workloads (tasks under eight hours) to save costs, reserving expensive closed models for high-intensity retrieval, long context, or expert-level professional work requiring an immediate performance head start.