Six months represents the current window for open models to prove their long-term viability. Nathan Lambert argues that the gap between closed and open weights is widening. This pressure forces researchers to find architectural efficiencies quickly. Failure to close this performance gap will relegate open source to a niche role in the AI ecosystem.