KPMG Report Highlights Kairos 3.1 as the Leading Global Model for Embodied Intelligence

iconKuCoinFlash
Share
AI summary iconSummary
KPMG released a report titled *Beyond the Singularity*, highlighting Kairos 3.1 as a leading global model for embodied intelligence. The report demonstrates that world models are central to AGI, enabling robots to perceive, predict, and act. Kairos 3.1, built on MetaEra, leads in real-world assets (RWA) news with strong performance in perception and execution. Industry trends show that 85% of world model funding deals in 2025 exceeded $10 million.
ME AI message: Recently, global professional services firm KPMG released a new report titled “Crossing the Singularity: How World Models Are Reshaping the Embodied Intelligence Industry and New Business Paradigms,” pinpointing a critical turning point in the development of embodied intelligence. The report asserts that large models alone cannot reach AGI; what truly enables robots to achieve a closed loop of “observe, reason, and act” is a world model capable of understanding physical laws, simulating future states, and guiding real-world actions. In its analysis of the world model industry landscape, technological pathways, and funding trends, KPMG identifies Daxiao Robotics as a representative of early-stage, high-valued enterprises. KPMG notes that Chinese world model companies are converging into two distinct development paths: one driven by star model teams and technological paradigms to secure capital valuation; the other leveraging embodied brains, industrial resources, and engineering-driven approaches to rapidly enter the unicorn tier—Daxiao Robotics is a prime example of the latter. This assessment from a third-party authoritative institution validates the leadership of Daxiao Robotics’ Kairos 3.1 world model within the field. Centered on an embodied brain, driven by real-world industrial scenarios, and anchored in an engineering闭环, Daxiao Robotics is accelerating the development of an embodied intelligence “ultimate brain” capable of understanding the world, simulating the future, and acting autonomously. Four Schools of World Models: Competition and Convergence Toward Integration The KPMG report identifies four dominant technical architectures in today’s global world model landscape. The first is representational world models, exemplified by JEPA and Dreamer, which primarily answer “What is the world?” by compressing environmental information into abstract representations of objects, space, states, and dynamic rules. The second is generative world models, represented by Sora and Genie, which focus on answering “What will the world become?” by generating images, videos, and scenarios to simulate potential future environments. The third is interactive world models, such as Marble and GAIA, which aim to answer “What happens if we change the world?” by treating actions as predictive conditions to simulate how different actions alter the environment. The fourth is integrated understanding-generating-predicting world models, including Kairos 3.1 and Cosmos 3, which strive to answer “How can we understand, predict, and autonomously change the world?” by unifying multimodal understanding, physical reasoning, future prediction, strategy planning, and action execution. These four schools are not simple replacements but are evolving in parallel across different dimensions. KPMG believes that world models will gradually converge toward integrated architectures: built on a unified world representation, enhanced with causal intervention capabilities, and dynamically equipped with offline strategy evaluation and real-time closed-loop control based on specific tasks. Therefore, evaluating the advancement of a world model should not rely solely on how realistic its generated visuals are—it must be assessed by its capacity for multimodal understanding, causal reasoning, dynamic evolution, future state prediction, strategy planning, and cross-scenario generalization. A truly advanced world model must enable robots not only to “see the world,” but to understand it, anticipate outcomes, and take correct actions. From L0 Generation to L5 Evolution: Kairos 3.1 Leads World Models Toward Action Intelligence KPMG summarizes the evolution of world model capabilities into six directions. For clarity, this progression can be further categorized into an L0-to-L5 capability ladder: L0: General Generation. Generates “possible futures” based on text, images, or video without guaranteeing physical realism or directly guiding robot actions. L1: Physical Consistency. Builds upon future state generation by incorporating understanding of physical laws—gravity, collision, contact, friction, motion, and deformation—to ensure predictions align with real-world constraints. L2: Causal Action Intervention. Treats planned robot actions as predictive conditions to simulate how different actions affect the environment, establishing causal links between “action—state—outcome.” L3: Offline Strategy Evaluation. Before actual execution, performs multi-step simulations of multiple complete action plans to compare task success rates, failure modes, time costs, and safety risks for informed strategy selection. L4: Real-Time Closed-Loop Control. Embeds the world model into the robot’s operational loop to continuously perform environmental perception, state prediction, action decision-making, task execution, and deviation correction with low latency. L5: Self-Evolution. Transforms real-world successes, failures, near-misses, and novel scenarios into reusable knowledge to continuously update the world model and action strategies—enabling capability transfer across tasks, scenarios, and robot forms. From L0 to L5, world models evolve from merely “generating content” to “understanding physics,” then to “predicting action consequences,” “guiding real-time control,” and finally “continuous self-evolution.” They are no longer just video models generating future scenes—they are becoming the cognitive core and action infrastructure of embodied intelligence systems. It is critical to note that L0 through L5 are not strictly linear upgrades; higher-level capabilities do not entirely replace lower-level ones. A truly advanced world model should be built on a unified world representation and dynamically integrate physical understanding, causal reasoning, strategy evaluation, and real-time control based on task requirements. Within this roadmap, Kairos 3.1 represents the most advanced direction in transitioning world models from generation to action. KPMG lists Daxiao Robotics as a representative enterprise rapidly entering the unicorn tier through embodied brains, industrial resources, and engineering-driven approaches. In its global mapping of leading world models, KPMG classifies Kairos 3.1 as a pioneer in perception and understanding, evolution and generation, and prediction and execution. This means Kairos 3.1 is not confined to single-task video generation or isolated control—it is continuously evolving toward an integrated architecture of understanding-generating-predicting-executing, anchored in L4 real-time closed-loop control and advancing toward L5 self-evolution. Its core objective is to enable robots to simulate the future before acting, correct in real time during action, and accumulate experience after action—ultimately forming a sustainable, self-evolving embodied intelligence “ultimate brain.” From L0’s “seeing the future,” to L1’s “understanding physics,” then L2–L3’s “simulating actions,” and finally L4–L5’s “real-time control and self-evolution,” Daxiao Robotics is driving world models across the critical inflection point from capability validation to industrial deployment—leading global world models toward true action intelligence. Kairos 3.1 Accelerates the Deployment of the Embodied Intelligence “Ultimate Brain” The KPMG report reveals that since late 2025, funding activity among world model companies has significantly accelerated. Among disclosed funding rounds, 85% exceeded $10 million USD, with 17 rounds surpassing $100 million USD. The surge in capital reflects a market shift from competition over individual robotic products to a battle for foundational world model capabilities and industrialization pathways. In its analysis of early-stage high-valuation companies, KPMG identifies Daxiao Robotics as a representative Chinese enterprise rapidly entering the unicorn tier through embodied brains, industrial resources, and engineering-driven approaches. As of June 30, 2026, Daxiao Robotics’ valuation in its Angel+ funding round reached industry-leading levels. Daxiao Robotics’ leadership stems from its refusal to confine world models to isolated algorithmic capabilities—instead focusing on building an embodied intelligence “ultimate brain” oriented toward the real world.On one hand, DaXiao Robot, centered on an embodied brain, aims to establish a unified understanding of the environment, objects, spatial relationships, and physical laws; on the other hand, it leverages world model technologies such as Kairos to enable robots to simulate future states before executing actions, comparing the success rates, risks, and efficiencies of different strategies. In KPMG’s embodied perspective mapping of leading international world models, Kairos is listed as a representative example across three directions: perception and understanding, evolution and generation, and prediction and execution. This demonstrates that DaXiao Robot is covering the complete evolution chain of world models, from cognition to action. The cutting-edge direction represented by DaXiao Robot is the transition from generative world models to control-oriented world models—enabling robots to understand the physical world and form stable environmental cognition; to simulate outcomes before acting, reducing trial-and-error costs; to perceive in real time, make dynamic decisions, and implement closed-loop control during execution; and to continuously learn from real-world operational feedback, achieving cross-task, cross-scenario, and cross-form transfer. This aligns closely with KPMG’s assessment of the future physical AI industry: world models will integrate multi-source data, multimodal representations, state prediction, action consequence simulation, and closed-loop validation, driving embodied intelligence from one-time delivery toward continuous training, deployment, and operation. Moving from generation-oriented to control-oriented, from “seeing the future” to “simulating the future,” world models are entering a decisive phase that will shape the industry landscape. With the embodied brain at its core and world model pathways like Kairos as its foundation, DaXiao Robot is advancing world models from laboratory capabilities to real-world industrial applications, building the “most powerful brain” of the embodied intelligence era and leading global world models toward higher-order action intelligence. (Source: Ifnar)
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.