千诀科技完成近亿元新一轮融资，类脑架构打造机器人"缸中之脑"

Core Insights - Qianjue Technology is developing a "brain in a jar" system that evolves across forms and scenarios, advancing embodied intelligence from single model capabilities to a system-level brain [1][3]. Financing and Growth - Qianjue Technology recently completed nearly 100 million yuan in Pre-A++ financing, with participation from six well-known investment institutions and existing shareholders. The funds will focus on the R&D of large models for embodied intelligent decision-making and planning, team building, and global commercialization [3]. - Since its establishment in June 2023, the company has completed seven rounds of financing, aiming to develop a universal intelligent base adaptable to various robotic hardware [3]. Unique Technological Approach - Unlike traditional "layered" and "end-to-end" approaches in embodied intelligence, Qianjue Technology employs a brain-like architecture that divides complex tasks into collaborative models for vision, hearing, decision-making, interaction, and memory. This allows for full-stack autonomy from chip to algorithm, enhancing flexibility and reliability for future system iterations and scene adaptations [4]. - The company has completed three generations of pre-trained "embodied brains," capable of autonomously executing a full loop of "perception - decision - action" without human prompts or pre-programmed rules [4]. "Brain in a Jar" Concept - The founder, Gao Haichuan, emphasizes that the company is not just creating an "intelligent brain" but a "brain in a jar" that can independently learn and evolve, allowing for rapid adaptation to different robotic forms. This decoupling of "brain" and "body" is crucial for addressing the challenges of diverse robotic forms and fragmented scenarios in the future [6][7]. Data and Engineering - Qianjue Technology has built the world's largest known dataset of pure real-world family scenarios, accumulating hundreds of millions of data points for embodied perception and decision-making training, covering various scenarios such as home, logistics, and industrial operations [9]. - The company has established an automated training data pipeline to lower the barriers to application, transforming traditional deep involvement from "secondary developers" to "direct users," significantly reducing the costs of robot application and adaptation [11]. Market Deployment Goals - The company has achieved comprehensive adaptation of its solutions to various mainstream robotic forms, including bipedal humanoid, wheeled, drone, quadruped, and vacuum robots, with several projects entering large-scale application phases. By 2026, Qianjue Technology aims to have over 100,000 robots equipped with its "embodied brain" [12]. - The self-developed Polibrain OS robot brain system is in the external validation phase, designed to integrate perception, decision-making, and control capabilities into a reusable and scalable universal intelligent base [12]. Addressing Industry Pain Points - Qianjue Technology's solutions address two core issues: the cost of environmental modification and the ability to respond to unexpected situations. Unlike traditional rule-based solutions that require significant physical modifications to deployment environments, Qianjue's autonomous decision-making model eliminates the need for such modifications [13]. - The largest current application scenario involves semi-service and household cleaning, such as hotel cleaning and public area maintenance, where the robots' core value lies in their autonomous decision-making capabilities rather than extreme dexterity [14]. Future Outlook - According to QYResearch, the global market for general brains in embodied intelligent robots is expected to grow at a compound annual growth rate of 52.0% from 2025 to 2031. The rapid growth of humanoid robots will continue to drive demand for universal intelligent bases [16]. - Qianjue Technology's unique brain-like partition architecture and "brain in a jar" concept are pushing the evolution of robotic intelligence from single model capabilities to a new stage of system-level brain development, potentially providing a replicable and scalable paradigm for the entire embodied intelligence industry [16].