發生了什麼事
Xieyue Intelligence, a startup co-founded by Chen Wei (former Li Auto chief scientist) and Zhang Xiao (former Li Auto product line president), has raised hundreds of millions of yuan in an Angel+ funding round. The capital, backed by investors including Linear Capital, PEAKVEST, Rhein Capital, and Hidden Hill Capital, will be used to train embodied foundation models, build infrastructure, and develop home robot hardware. This is the company's second funding round in seven months, following an initial Angel round led by Li Auto and Vision Plus Capital.
Gasgoo reports that Xieyue Intelligence has secured hundreds of millions of yuan in an Angel+ funding round. The investors include Linear Capital, PEAKVEST, Rhein Capital, and Hidden Hill Capital. The funds are designated for training embodied foundation models, building computing and data infrastructure, expanding the core team, and developing home robot hardware for scenario verification.
The startup was co-founded by Chen Wei, former chief scientist of AI and head of the department at Li Auto, and Zhang Xiao, the automaker's former product line president. The core team includes members from global tech, AI, robotics, and intelligent automotive firms. Gasgoo notes the team has experience training large models across 10,000 GPU cards and managing mass-production chains for consumer products.
Strategically, Xieyue Intelligence is positioning the home as the core training ground for its embodied foundation models, diverging from the industry's focus on industrial and logistics scenarios. The company is adopting a gradual strategy to prioritize semi-structured environments like hotels and nursing homes to prove model capabilities and unit economics before moving into household settings. Initial targets include high-frequency chores such as laundry, organizing, and cleaning.
The company is building a closed-loop ecosystem encompassing robot hardware, embodied foundation models, and home-based self-evolution. This system utilizes 'Duplex Reasoning' for the , a 'Human-centric' approach for data collection, and a 'Safety-first' framework. The hardware strategy involves a 'single body, full-stack closed-loop' approach to control variables initially before rolling out consumer products.
為什麼這很重要
The launch signals a strategic shift in the embodied AI sector, with a high-profile team from the automotive industry pivoting to consumer home robotics. By focusing on 'Duplex Reasoning' and a 'verify first, enter home later' strategy, the company aims to address the complexity of household environments, which are considered critical testbeds for general embodied intelligence. This move highlights the growing convergence of automotive AI expertise and consumer robotics.
This funding round represents a significant entry into the embodied AI space by a team with deep expertise in automotive AI and large-scale model training. The shift from industrial to home-focused robotics addresses a complex, high-frequency physical environment that is considered a key touchstone for validating general embodied intelligence.
The company's technical approach, specifically 'Duplex Reasoning,' aims to maintain a two-way channel between the robot and human during perception, reasoning, and execution. This allows for dynamic adjustment of action goals and interruption of commands, which is critical for safe and effective operation in unstructured home environments.
By focusing on high-quality data and systematic infrastructure rather than just larger models, Xieyue Intelligence is attempting to differentiate itself in a crowded market. The company argues that core competitiveness in embodied intelligence lies in data quality and infrastructure capabilities, which are viewed as equally critical pillars to model size.
互動機制：它實際上是如何運作的
以互動方式探索這項發展背後的基礎技術。
An agent must create a draft calendar event for Tuesday at 2 p.m. Which evidence would establish the requested result?
接下來看什麼
Monitor the company's progress in semi-structured environments like hotels and nursing homes, the operational status of its data by 2026, and the eventual release of consumer-facing home robot products.
The company plans to have its full data —from collection, cleaning, and labeling to training—fully operational by 2026. Progress in this area will be a key indicator of its ability to scale its embodied foundation models.
Success in semi-structured environments like hotels and nursing homes will be a prerequisite for the company's entry into household settings. Monitoring the unit economics and model capabilities in these initial deployments will provide insight into the viability of its 'verify first, enter home later' strategy.
The development and release of consumer-facing home robot products will be the ultimate test of the company's 'single body, full-stack closed-loop' hardware strategy and its ability to manage the complexity of real-world home environments.