Spirit AI needed stable high-performance infrastructure to train embodied AI models efficiently enough for commercial robotics, especially because flexible-object manipulation, long-sequence multimodal training, and rapid model iteration make embodied intelligence unusually compute-intensive.
Reported impact
Spirit AI gained faster and more efficient embodied-model development, improving the practicality of deploying robotics systems in complex real-world scenarios such as manufacturing and service operations.
Baidu AI Cloud provided Spirit AI with an AI Infra stack tuned for embodied-model development so the company could train and iterate its Spirit v1 vision-language-action model efficiently. Spirit AI's core challenge was not only building a model 'brain' for high-level instruction understanding and task planning, but also handling the 'cerebellum' motion-control layer, embodied datasets, and the supporting software and hardware system. Baidu's contribution focused on making cloud compute productive: a stable high-performance AI Infra platform backed Spirit's training jobs, while an associated training-and-inference acceleration suite improved throughput and reduced wasted compute. The platform is also being optimized for long-sequence performance and multimodal training, which matters because Spirit v1 must fuse language, scene perception, and robot action control to solve flexible-object tasks such as folding clothes. This let Spirit move faster from research to commercially relevant embodied-AI scenarios without having to build the whole training substrate from scratch.
Natural-language instructions and multimodal environment data are fed into Spirit v1 during cloud training on the AI Infra platform, accelerated training pipelines optimize the model for dynamic-scene perception and action control, and the resulting VLA model drives downstream embodied-AI tasks such as long-horizon flexible-object manipulation.
Baidu AI Cloud is also optimizing long-sequence performance and multimodal training architectures for embodied large-model development. (original: 针对具身大模型开发的差异化特点，百度智能云也正着力于优化长序列性能和多模态训练架构。)
Baidu AI Cloud's embodied-AI platform can raise effective cluster training time to 98.8% and its large-model acceleration suite improves training and inference efficiency by 30% and 60%. (original: 百度智能云具身智能AI Infra技术平台，可帮助用户将集群有效训练时长提升到98.8%...配套的大模型训推加速套件也分别大幅提升训练和推理效率30%和60%。)
Spirit AI aims to commercialize embodied intelligence in trillion-yuan markets such as intelligent manufacturing and services. (original: 未来，百度智能云将全力支持千寻智能...并在智能制造、服务行业等万亿级市场推进商业化落地。)
Spirit AI used Baidu AI Cloud's high-performance AI Infra platform to train the Spirit v1 VLA model efficiently. (original: 百度智能云为千寻智能提供了高性能稳定可靠的AI Infra基础设施，支持VLA视觉语言动作模型高效训练。)
Spirit v1 solved long-horizon flexible-object manipulation by using dynamic-scene perception and precise temporal action control so the robot could follow natural-language instructions to fold clothes and do other complex tasks. (original: 该模型通过动态场景感知技术，解决了衣物褶皱、形态随机性带来的操作难题，并在时序控制上实现高精度动作策略调整，使机器人能够通过自然语言指令完成叠衣服等复杂任务。)
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Baidu AI Cloud Sample Dec 29, 2025 Beijing Yilian Tongda Technology Co., Ltd. Yilian Tongda needed a safer and more efficient way to help power-industry staff work with scattered standards, maintenance records, inspection documents and risk-control knowledge that were difficult to search and keep… Agent Technology · Qianfan AppBuilder · Qianfan ModelBuilder Baidu AI Cloud Sample Dec 29, 2025 Hualvwang Hualvwang needed a way to help ordinary users describe legal issues more precisely, retrieve relevant legal knowledge faster and scale personalized legal-service triage across a very large user base. Daily ERNIE invocation peaks exceeded 10 million. Qianfan AppBuilder · Qianfan ModelBuilder · Recommended Questions Baidu AI Cloud Sample Dec 29, 2025 Shanghai Xinzhi Software Co., Ltd. (上海新致软件股份有限公司) Shanghai Xinzhi Software needed a scalable way to help automotive sales teams cope with high staff turnover, expensive talk-track training, fragmented customer-profile information, and weak use of historical call data… Average sales score increased 25%. Baidu Search plugin · ERNIE-4.0-Turbo · Qianfan AppBuilder 