
小米发布了小米机器人-1作为开源项目———一个通用型嵌入式AI基础模型,而非仅针对小米自家机器人。如今,在Nvidia GR00T专有模型与Hugging Face社区打造的LeRobot开源模型的对比下,一个资源充足的开放权重替代方案应运而生。
简单来说: 小米发布了其机器人AI基础模型的权重供所有人使用。目标是让它成为下一代机器人的核心支柱——而不仅仅是小米自己的。
据TechNode报道,小米开源了小米机器人-1(Xiaomi-Robotics-1),这是一个具身AI基础模型。此次发布涵盖了从真实机器人后训练到模型部署的全流程,并包含用于基准评估的代码。该模型被定位为通用型机器人应用的基础支柱,面向不同制造商,遵循开放权重的基础层策略,以加速小米硬件生态系统的采用。
这是将Meta/Llama策略应用于机器人的典范。小米押注开源基础模型能在其硬件上创造生态系统依赖——正如Meta通过Llama推动开发者采用其平台一样。时机恰到好处:Nvidia的GR00T N1.2仍为专有,Hugging Face的LeRobot由社区构建且资源有限,而小米现在提供了一个经过生产验证、资源充足且开放权重的替代方案。具身AI竞争如今真正呈现多极化态势——而开源轴心可能决定生态系统将构建在哪个硬件栈之上。
第三方机器人制造商(尤其是中国以外的)是否会采用小米机器人-1,以及它在标准操作任务基准测试中与GR00T N1.2的对比表现。
本文由人工智能撰写,并经人工编辑审核。
Open-sourcing is great, but will Xiaomi-Robotics-1 actually translate to affordable, consumer-ready robots? Open models alone don’t guarantee accessibility.
After years of Nvidia’s GR00T monopolizing robotics AI stacks, this open-source alternative could really shake things up. Wonder how quickly developers will jump on it.
GR00T’s hardware integration might still give it an edge, but if Xiaomi’s model scales well on commodity devices, devs could actually shift faster than expected.
Still skeptical about open-source robotics models catching up to proprietary ones. Time will tell if this delivers real-world reliability.
Open-source models often improve faster precisely because they rely on collective debugging-proprietary ones can’t afford that same scrutiny.
Xiaomi-Robotics-1 could democratize robotics if the community rallies fast-open-source models scale faster when they avoid vendor lock-in.
I wonder if open-sourcing this model will accelerate real-world applications faster than GR00T’s curated access. Licensing terms will make or break adoption.
Xiaomi-Robotics-1 feels more like a strategic push than a technical leap - open source shouldn't just compete with GR00T but redefine what robotics needs at the hardware layer too.
Course aux modèles fondation embodied : X-Square, Xiaomi, GR00T