在 2026 年 2 月 18 日，一项突破性测试让 1 辆地面车辆与 2 架无人机在 500-meter 范围内协同运作，展示了自主系统的进展。此次演示凸显了由神经网路模型规模差异所带来的显著改进，其中一些模型超过 100 billion 参数，另一些约为 10 billion。约 10:1 的规模差距，突显了为国防应用开发更大、更复杂 AI 模型的趋势。

这些发展源自 Department of Defense 的 4 份合约，目标是推进无人系统中人工智慧自主性与协同能力的边界。这些合约持续 1 年或更久，支持了研究与部署工作，并正在扩大自主车辆与无人机的普及。整合不同规模的模型，使多平台上的效能与适应性得以最佳化。

虽然有关作业成功的精确统计仍然有限，但其意涵显示，正日益重视将大规模 AI 与更小、更高效率的模型结合，以实现多用途且稳健的自主系统。这种方法反映了国防技术的更广泛趋势：在管理计算资源的同时提升自主能力。持续进行的努力显示，在各种作业情境中，自主系统正强力走向更精密、更可靠且更广泛的使用。

"On February 18, 2026, a groundbreaking test involved one ground vehicle and two drones operating collaboratively within a 500-meter range, showcasing advances in autonomous systems. This demonstration highlighted significant improvements fueled by neural network models varying greatly in size, with some exceeding 100 billion parameters and others around 10 billion. The scale difference, roughly a 10:1 ratio, underscores the trend toward developing larger, more complex AI models for defense applications.",

"The developments stemmed from four Department of Defense contracts aimed at pushing the boundaries of artificial intelligence autonomy and coordination in unmanned systems. These contracts, spanning a year or more, supported research and deployment efforts that are expanding the proliferation of autonomous vehicles and drones. The integration of models of diverse scales allowed for optimized performance and adaptability across multiple platforms.", (Key numbers: 4, 1)

"While exact statistics about operational success remain limited, the implications suggest a growing emphasis on combining large-scale AI with smaller, more efficient models to achieve versatile and robust autonomous systems. This approach reflects a broader trend in defense technology focused on increasing autonomy capabilities while managing computational resources. The ongoing efforts indicate a strong trajectory toward more sophisticated, reliable, and widespread use of autonomous systems in various operational contexts."