Tencent Robotics introduced three embodied fundamental models during the WAIC 2026 event. These models include a VLM for scene understanding, the cognitive model RxBrain for planning considering visual states, and VLA for continuous actions at a frequency of 500–1000 Hz.
Chief Scientist Dr. Zhang Zhengyou presented this architecture as a solution to a fundamental problem that current VLA approaches do not adequately address. He noted that different types of robotic intelligence must operate at different frequencies because the physical world itself operates across multiple timescales.
Understanding this issue arose after a failed experiment transferring robot capabilities to OpenClaw, which caused a delay of tens of seconds—unacceptable for physical robots where even a cognitive delay of 2–3 seconds is problematic.
The three-layer architecture solves this problem by separating frequencies. The Hy-Embodied-VLM model is responsible for spatial understanding at a lower frequency, interpreting visual scenes and object relationships. Hy-Embodied-RxBrain represents a significant innovation in cognitive architecture: it not only generates task steps in language but also visualizes how the physical world should look after each subtask is completed, using visual state representations instead of textual descriptions.
This eliminates the main limitation of language-only planning systems, as textual descriptions cannot effectively convey spatial relationships and physical constraints that can be instantly communicated through vision. The Hy-Embodied-VLA model operates at the highest frequency, translating high-level goals into continuous low-level action commands at a speed sufficient for real-time interaction with the physical world.
Tencent reported that Hy-Embodied-VLA underwent production testing at a chemical plant. The model handles highly mixed, low-batch, SKU-intensive assembly lines, achieving a cycle time of less than 6 seconds per part. Furthermore, the model can adapt to new SKUs in just 8 hours of data collection and fine-tuning.
The testing conditions differ significantly from laboratory demonstrations: objects are placed randomly, lighting changes, production interruptions occur, and the robot must achieve an almost zero failure rate. Dr. Zhang emphasized that a demonstration scoring 80–90 points without actual deployment holds little value, and factory testing reflects a commitment to measurable industrial performance rather than benchmarking metrics.
The RxBrain cognitive model addresses a second complex challenge—the linguistic bottleneck in robotic reasoning. Previous Tairos systems relied on text for inter-brain communication, but describing object spatial relationships in words can be verbose and ambiguous. By incorporating a visual state representation into the planning loop, RxBrain ensures higher information throughput between perception and action.
The model is described as an embodied cognitive world model, not a complete world model, as a consensus architecture for a unified world model has not yet been reached in this field. Together, the three models and the Tairos platform form Tencent's complete embodied intelligence stack, available for any developer or manufacturer to use. Thus, the robotics community can now build upon the same frequency-sensitive architecture developed by Tencent through real failures and iterative refinement.
LimX Dynamics has demonstrated how the full-sized humanoid robot Oli is capable of independently performing prolonged household chores in a real living space without requiring remote control or editing.
The robot, featuring 31 degrees of freedom, continuously performed a series of household duties in one uninterrupted cycle, including folding clothes, organizing items, stacking boxes, collecting trash, and delivering water. This achieved level of long-term autonomous manipulation places LimX second in the world, alongside the company Figure.
The breakthrough is attributed to LimX's COSA 0.5 brain architecture, which consciously rejects the commonly accepted industry model for brain function. Founder Zhang Wei asserts that equating a large model with brain functions is erroneous. He compares a pure large model to Stephen Hawking lying in bed: extremely intelligent but completely incapable of movement. In Zhang Wei's view, the brain is a system that organizes cognitive abilities, skills, and motor control into a single architecture operating on a real physical body.
The COSA system implements a three-stage architecture, S2-S1-S0, functioning across different time scales. The S2 Cognitive level acts like the prefrontal cortex, utilizing an LLM/VLM agent for scene understanding, memory, world modeling, reasoning, and human interaction. This level determines which actions need to be performed. The S1 Skills level provides a set of trained capabilities, including Vision-Language-Action (VLA) models for generating full-body motion. At the S0 Motion Control level, it works with a 10-million-parameter Whole Body Trajectory (WBT) policy at a frequency of 1000 Hz directly on the device, translating arbitrary target movements from higher levels into balanced, coordinated joint commands.
These layers interact through intentionally narrow interfaces: intent is passed down, and the robot's state is passed up asynchronously, without blocking operation. This architecture allows for stable iteration, something approaches based on a single model cannot achieve. Each layer can be updated, replaced, or fine-tuned independently without affecting the others. In the Oli demonstration, reinforcement learning was used on the real robot, where experts remotely corrected errors, while S0 maintained balance autonomously. Collected data was used to train reward models, and RL continued to iterate on the real robot, making the system stronger with each use, rather than stopping after initial training.
The home demonstration also confirmed LimX's strategic focus on out-of-factory applications. While most embodied AI companies focus on industrial deployment, LimX CEO Zhang Wei is orienting humanoids toward commercial service, hospitality, entertainment, and domestic care markets, where value is created through physical interaction similar to human interaction. The possibility opened up by COSA 0.5 makes this strategy increasingly plausible, demonstrating that a system designed for consumer applications can achieve world-class performance in autonomous manipulation.
