AI热点 1周前 81 阅读 0 评论

AgiBot Unveils World Model Platform and Eyes Global Expansion at WAIC 2025

作者头像
AI中国

AI技术专栏作家 | 发布了 246 篇文章

TMTPOST -- AgiBot, a rising force in embodied AI, unveiled major advancements in dual-arm robotics and TMTPO systems at the 2025 World Artificial Intelligence Conference (WAIC), as the Shanghai-based firm doubles down on its push toward real-world deployments and international expansion.

At the conference on July 27, AgiBot Partner and Embodied Intelligence Division President Yao Maoqing introduced “Genie Envisioner” (GE), an open-source world model platform tailored for dual-arm robots. Touted as an industry first, the system integrates prediction, control, and evaluation, offering an all-in-one framework to take robots from perception to execution.

Backed by proprietary data infrastructure, AgiBot also revealed AgiBot World, billed as the world’s largest real-world dataset for embodied AI. The company plans to release its next-generation dual-arm robot, Genie G2, with improvements in motion precision, scenario adaptability, and onboard data collection capabilities, aimed at industrial, commercial, and home use cases.

Luo Jianlan, Chief Scientist of Turing Smart Robotics

AgiBot is scaling quickly. It recently won a 78 million yuan contract from China Mobile to supply humanoid robots for customer-facing tasks in service halls. These bipedal units are also being prepped for deployment in hotels, banks, and other high-touch environments.

Chief Scientist Luo Jianlan noted that the firm’s multimodal large models are now trained entirely on real-machine data, a shift she said reflects the industry’s broader move from “demo-first” to full-stack commercialization.

He emphasized AgiBot’s focus on safety, scenario migration, and human-friendly design, distancing the company from high-profile robot malfunctions she attributed to poor engineering rather than autonomy risk.

Yao Maoqing, Partner and President of the Embodied Intelligence Division at AgiBot

Founded in early 2023, AgiBot has quickly become one of China’s most closely watched AI startups. Its co-founder and CTO, Peng Zhihui—known online as “Zhihuijun”—was a former “Huawei Genius Youth” and a widely followed figure in the hardware and open-source community.

In July, the company took a key step toward public listing. It moved to acquire a majority stake in Swancor Advanced Materials in a 2.1 billion yuan deal that would bring AgiBot to Shanghai’s STAR Market. With investors including Tencent, Baidu Ventures, Sequoia China, JD Technology, BYD, and Hillhouse Capital, AgiBot’s latest valuation reportedly exceeds 26 billion yuan.

Beyond humanoids, AgiBot is continuing to iterate its Lingxi quadruped series, with the new D1 Ultra targeting applications where cost, reliability, and terrain versatility are priorities. According to Yao, the product line shares components with humanoid systems, enabling scale efficiencies across the firm’s expanding hardware portfolio.

On July 26, Peng unveiled the company’s layered open-source robotics software stack—Lingqu OS—designed to accelerate industry adoption. The framework includes a real-time middleware layer (AimRT), a standardized intelligent agent service layer, and a toolchain for simulation, training, and deployment. The OS will begin rolling out in Q4 under an open co-development model.

With operational footprints already in North America, Europe, the Middle East, and Asia-Pacific, AgiBot is pushing ahead with its “globalization through localization” strategy. Yao noted that while China remains hypercompetitive, overseas markets offer fertile ground for AI-native hardware to scale.

“Getting robots into factories is only the beginning,” Yao said. “Real traction will come from everyday environments—retail, service, education, and the home.”

作者头像

AI前线

专注人工智能前沿技术报道,深入解析AI发展趋势与应用场景

246篇文章 1.2M阅读 56.3k粉丝

评论 (128)

用户头像

AI爱好者

2小时前

这个更新太令人期待了!视频分析功能将极大扩展AI的应用场景,特别是在教育和内容创作领域。

用户头像

开发者小明

昨天

有没有人测试过新的API响应速度?我们正在开发一个实时视频分析应用,非常关注性能表现。

作者头像

AI前线 作者

12小时前

我们测试的平均响应时间在300ms左右,比上一代快了很多,适合实时应用场景。

用户头像

科技观察家

3天前

GPT-4的视频处理能力已经接近专业级水平,这可能会对内容审核、视频编辑等行业产生颠覆性影响。期待看到更多创新应用!