{"id":18966,"date":"2025-09-14T17:49:07","date_gmt":"2025-09-14T16:49:07","guid":{"rendered":"https:\/\/visegradpost.com\/?p=18966"},"modified":"2025-09-13T15:47:46","modified_gmt":"2025-09-13T14:47:46","slug":"chinas-brain-like-ai-beats-chatgpt-by-100x-spikingbrain-processes-4-million-tokens-using-20-watts-on-metax-chips","status":"publish","type":"post","link":"https:\/\/visegradpost.com\/en\/2025\/09\/14\/chinas-brain-like-ai-beats-chatgpt-by-100x-spikingbrain-processes-4-million-tokens-using-20-watts-on-metax-chips\/","title":{"rendered":"“China’s Brain-Like AI Beats ChatGPT by 100X”: SpikingBrain Processes 4 Million Tokens Using 20 Watts on MetaX Chips"},"content":{"rendered":"
\n\n\n\n\n
IN A NUTSHELL<\/strong><\/td>\n<\/tr>\n
\n
    \n
  • \ud83e\udde0 Researchers in China introduce SpikingBrain 1.0<\/strong>, a groundbreaking AI system mimicking the human brain’s efficiency.<\/li>\n
  • \u26a1 Utilizes spiking computation<\/strong> to achieve faster processing and reduced energy consumption compared to traditional models.<\/li>\n
  • \ud83d\udcbb Operates stably on MetaX chips<\/strong>, highlighting potential for future real-world applications without relying on Nvidia hardware.<\/li>\n
  • \ud83c\udf0d Advances in neuromorphic computing<\/strong> could transform industries by making AI technologies more sustainable and accessible.<\/li>\n<\/ul>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/figure>\n

    In the ever-evolving landscape of artificial intelligence, a groundbreaking development has emerged from China. Researchers at the Chinese Academy of Sciences\u2019 Institute of Automation in Beijing have unveiled SpikingBrain 1.0, a novel AI system designed to emulate the efficiency of the human brain. This model stands out for its ability to handle extensive data strings while operating on significantly lower power. By leveraging homegrown MetaX chips instead of relying on industry-dominant Nvidia hardware, SpikingBrain 1.0 represents a major leap in AI technology. The system\u2019s remarkable performance and efficiency promise to open new avenues in various fields, from healthcare to legal research.<\/p>\n

    Spiking Computation: A Game-Changer<\/h2>\n

    At the heart of SpikingBrain 1.0 is the innovative approach of \u201cspiking computation.\u201d This method mimics the way biological neurons operate by firing signals only when triggered by specific inputs. Unlike traditional AI models that activate entire networks to process data, SpikingBrain 1.0 remains largely inactive until necessary, leading to significant energy savings and faster processing times. This selective response reduces energy consumption and accelerates data handling, making it a more efficient alternative to mainstream AI tools like ChatGPT.<\/p>\n

    To illustrate its capabilities, researchers developed two versions of the model. One features 7 billion parameters, while the other boasts 76 billion. Both versions were trained using a total of approximately 150 billion tokens, a relatively modest amount for models of this scale. The system excelled in tests requiring the handling of long data sequences, outperforming conventional models by a wide margin. For instance, in one scenario, the smaller model processed a 4 million token prompt over 100 times faster than standard systems.<\/p>\n

    \u00ab\u00a0100 Megawatt Molten Salt Reactor\u00a0\u00bb: Korean Engineers Create World’s First Nuclear Powered LNG Carrier Eliminating Lifetime Fuel Replacement<\/a><\/p><\/blockquote>\n