ACE ROBOTICS announced today that its open-source Kairos world model has achieved leading results across four global embodied-intelligence benchmarks: RoboTwin 2.0, LIBERO-Plus, WorldModelBench Robot and DreamGen. As of June 12, 2026, Kairos ranked first among evaluated world models and vision-language-action (VLA) systems on these benchmarks' public leaderboards, leading across core capabilities including complex robotic manipulation, scene-level generalization, physical-world modeling and zero-shot transfer. The project is openly available on GitHub, Hugging Face and ModelScope, providing a public reference point for researchers and developers.
The significance of these results lies in the fundamental challenge of embodied intelligence: generalization. Robots must operate reliably in unseen environments, adapting to new lighting, layouts, objects, embodiments and noisy real-world conditions. While VLA models have become prevalent by mapping perception and language inputs directly to actions, ACE ROBOTICS believes world models offer a more scalable path by learning the underlying dynamics of the physical world and predicting how environments evolve. Kairos is designed to validate that approach, and its benchmark performance suggests world models may be a superior foundation for general-purpose robots.
On the LIBERO-Plus benchmark, which evaluates robustness under seven real-world variables, Kairos achieved an overall score of 89.0, surpassing leading VLA models including ACoT-VLA (88.0) and Pi 0.5 (85.7). This marks the first time a world-model approach has outperformed VLA systems on scene-level generalization, pointing to a future where robots adapt to homes, factories and retail spaces with far less environment-specific retraining. Kairos showed near-ceiling performance on lighting (97.7), noise (96.8) and background (95.8).
On WorldModelBench Robot, Kairos-4B, with only 4 billion parameters, achieved an overall score of 9.30, outperforming larger systems including 28-billion-parameter Lingbot and 16-billion-parameter Cosmos 3. It matched the top instruction-following score of Cosmos 3 with about one quarter of the parameters, a fourfold efficiency gain. ACE ROBOTICS attributes this to its native unified architecture, which integrates understanding, generation and prediction within a single backbone, reducing information loss and coordination latency. NVIDIA's Cosmos 3.0, introduced in 2026, adopts a comparable single-system design, indicating industry convergence.
Kairos also ranked first on DreamGen Bench for synthetic data transfer, and on RoboTwin 2.0 it scored 96.1%, achieving state-of-the-art results on complex dual-arm manipulation tasks. Together, these results validate Kairos across the core dimensions of embodied intelligence, supporting ACE ROBOTICS' aim to move robots beyond task imitation toward physical-world understanding and real-world execution.
The results come as ACE ROBOTICS accelerates commercialization, having raised several hundred million U.S. dollars in the first half of 2026, with backing from investors such as Dachen Caizhi, Shenzhen Capital Group and the Shanghai Sci-Tech Innovation Fund. The proceeds will support continued world-model research and integrated hardware-software solutions for smart retail, security, tourism and hospitality. 'Embodied intelligence is the next era of AI, and a world model is the key to unlocking it,' said Wang Xiaogang, Chairman of ACE ROBOTICS. 'Our mission is to give every robot a capable brain.'
Kairos is openly available on GitHub, Hugging Face and ModelScope.
