The current window for open-source AI viability is narrowing. Nathan Lambert argues that open models face a critical survival test over the next six months. This pressure stems from the widening gap between proprietary systems and open weights. Practitioners must now determine if open research can keep pace with closed labs.