Jakarta, INTI – 9, September, 2025 Amid the global race to develop more efficient and energy-saving artificial intelligence (AI) technologies, a research team from China has unveiled a breakthrough that mimics the workings of human brain neurons. This innovation is expected to pave the way for next-generation computing with lower power consumption, while challenging the dominance of Transformer-based AI models such as ChatGPT.
AI Breakthrough Based on Spiking Neurons
Scientists from the Institute of Automation, Chinese Academy of Sciences (CAS), have launched SpikingBrain-1.0, a large-scale AI model fully trained on domestically developed GPUs. Unlike mainstream AI systems that require massive datasets and computing power, this model employs a spiking neuron approach that is significantly more energy-efficient.
Using only about 2 percent of the pre-training data typically required by conventional large models, SpikingBrain-1.0 matches the performance of several open-source models in language comprehension and reasoning tasks.
High Efficiency with Minimal Data
SpikingBrain’s advantages go beyond energy savings, extending to speed efficiency. During the inference stage, one variant of the model demonstrated a 26.5-fold increase in speed compared to Transformer architectures when generating the first token from a one-million-token context.
Its ability to handle ultra-long sequences makes it highly valuable for applications such as legal and medical document analysis, high-energy particle physics research, and DNA sequence modeling.
Potential for Neuromorphic Hardware
Xu Bo, Director of the CAS Institute of Automation, emphasized that this innovation opens a non-Transformer pathway for the future of AI development. “This model has the potential to inspire the design of next-generation neuromorphic chips with far lower power consumption,” he said.
The research team has also released the model as open-source, providing a public testing platform along with a bilingual technical report validated by industry experts.
Innovation Legacy: The Energy-Efficient “Speck” Chip
Previously, in 2024, CAS scientists in collaboration with Swiss colleagues developed the neuromorphic chip Speck, designed to mimic neuron and synapse functions in the human brain. The chip consumes just 0.42 milliwatts when idle, making its energy use virtually negligible.
For comparison, the human brain with billions of neurons operates on just around 20 watts of power, far more efficient than today’s Transformer-based AI systems.
Conclusion
The introduction of SpikingBrain-1.0 marks a major step forward for China in delivering next-generation AI technology that is efficient, energy-saving, and environmentally friendly. By adopting a brain-inspired approach, this innovation is expected to revolutionize AI development while accelerating the rise of future neuromorphic hardware.
Read More:Komdigi Trains Civil Servants to Develop Strategic AI Solutions Through AI Connect Academy