Chinese AI Firm Zhiping Unveils Brain-Inspired NeuroVLA Model at UN, Offering a Low-Power Alternative to Compute-Intensive Robotics AI

As the global AI industry continues its arms race of stacking computing power, data, and energy, a Chinese company from the Guangdong-Hong Kong-Macao Greater Bay Area chose the United Nations headquarters in New York to present a fundamentally different approach. At the UN Open Source Week, Kristine Mo, Vice President of Overseas and Ecosystem at Zhiping Technology, shared the stage with Turing Award winner Yann LeCun to demonstrate NeuroVLA, billed as the world's first original brain-inspired large model, alongside the one-stop open-source AlphaBrain Platform.

"Robot hardware is advancing rapidly, but the real bottleneck remains intelligence," Mo said during her speech. The current predicament is that large models have an almost insatiable appetite for data and electricity. Zhiping founder and CEO Guo Yandong previously stated at the Summer Davos that if everyone continues along the same large model trajectory, the industry would need 10 times more data and 10 times more power, yet the real world does not have infinite data and energy.

NeuroVLA draws inspiration from the human brain's "cortex-cerebellum-spinal cord" three-layer collaborative mechanism, enabling learning and decision-making with far less data. The model reduces robot motion jitter by over 75%, and the spinal layer's average power consumption is as low as 0.4 watts, breaking the high-power, high-latency constraints of traditional VLA models.

Choosing the UN as the stage for this demonstration signals Zhiping's ambition to offer a "Chinese solution" for global AI development: achieving intelligent leaps through architectural innovation rather than resource accumulation. The company positions this as not only a technological breakthrough but also a more sustainable direction for the path toward general intelligence.

Zhiping is no small startup. After completing seven funding rounds, the company has gone through five more series B rounds, raising over 1 billion RMB with a valuation exceeding 10 billion, signaling strong capital market confidence in its brain-inspired approach.

The unveiling comes as the embodied AI sector accelerates globally, with industry giants like Nvidia open-sourcing robotic skill libraries. Zhiping's differentiated strategy of finding the "robot brain's" evolutionary code from human neural mechanisms offers a compelling alternative to the Western compute-centric paradigm.

The key question ahead is whether NeuroVLA's open-source release can attract a substantial developer ecosystem and whether its low-power architecture can consistently validate its theoretical advantages in real-world deployments.