On March 14th, leading embodied intelligence innovator AgiBot held a press conference in Zhuhai to announce the launch of the world's first generalist embodied foundation model, the Genie Operator-1, along with the fully intelligent agile humanoid robot Lingxi X2. In collaboration with Huafa Group and other ecosystem partners, the company also officially initiated four industry-specific models. The unveiling of Genie Operator-1, an independently developed generalist embodied foundation model, marks a major leap toward greater generalization, openness, and intelligence in embodied AI.

On-site event

The Genie Operator-1 introduces the pioneering ViLLA (Vision-Language-Latent-Action) framework, which integrates VLM (Vision-Language Model) and MoE (Mixture of Experts). This novel approach enables human-like video learning, few-shot generalization, cross-embodiment adaptation, and continuous self-evolution. Leveraging vast datasets from both humans and various robots, the model grants robots revolutionary learning capabilities—allowing them to generalize across different environments and objects, quickly adapt to new tasks, acquire new skills, and deploy across diverse robotic platforms while continuously evolving.

Peng Zhihui (the founder of AgiBot) showcases Lingxi X2 in Zhuhai

Harnessing the fine-tuning capabilities of Genie Operator-1 and Zhuhai's specific industry data, AgiBot has joined forces with multiple partners to launch four sector-specific models, spanning smart housing, healthcare, security, and retailing.

The Lingxi X2, unveiled at the event, is a general-purpose humanoid robot that integrates advanced mobility, interaction, and operational intelligence. Featuring high-degree-of-freedom movement and an affective computing engine, it is capable of executing a wide range of general tasks, embodying the seamless fusion of AI and humanoid robotics.

On-site event

As a benchmark enterprise in AI and robotics innovation, AgiBot has established a cutting-edge full-stack "Ontology + AI" technology framework. Its three major robotic families—Yuanzheng, Jingling, and Lingxi—have already surpassed 1,000 units in mass production.

Source: Lingnan on the Cloud

全球首个通用具身基座模型在珠海发布

3月14日，具身智能领军者智元机器人在珠海举办发布会，发布全球首个通用具身基座模型智元启元大模型（Genie Operator-1）与全智能灵动机器人灵犀X2，并联合华发集团等生态链合作伙伴，正式启动四大行业模型。发布会上，智元机器人正式发布了自主研发的全球首个通用具身基座模型——智元启元大模型（Genie Operator-1），标志着具身智能向通用化、开放化、智能化方向快速迈进。

智元启元大模型开创性地提出了ViLLA框架，该框架由VLM(多模态大模型) + MoE(混合专家)组成，将人类视频学习、小样本快速泛化、一脑多形、持续进化等突出优势集于一身。通过人类及多种机器人数据，它赋予机器人革命性学习能力，可在不同环境物品中泛化应用，快速适应新任务、学习新技能，还能部署不同机器人本体并持续进化。

基于智元启元大模型的精调能力与珠海特定场景数据，智元机器人联合多方启动“科技+好房子、康养、安防、泛零售”四大行业模型。

现场亮相的灵犀X2是通用人形机器人，兼具运动、交互和作业智能，支持高自由度运动，搭载情感计算引擎，能执行通用任务，完美融合人工智能与人形机器人技术。

作为“人工智能+机器人”融合创新标杆企业，智元机器人构建了领先的“本体+AI”全栈技术，旗下远征、精灵、灵犀三大机器人家族，已有超1000台通用具身机器人量产下线。

文｜记者 钱瑜
图｜受访者供
翻译｜杨钧雅
审校｜赵凡