The arrival of DeepSeek has further accelerated the existing global race for artificial intelligence. Several companies have carefully analyzed the success of Chinese AI to develop new, even more modern tools.
This is the case of the startup MiniMax, also from China. It has just launched a new model capable of performing complex productivity tasks and that, according to the company, surpasses all competitors in the country.
New AI can process more information simultaneously
In several performance tests, the MiniMax-M1 model obtained higher scores than the latest version of DeepSeek, the R1-0528. According to the startup, M1 supports a context length of one million tokens, eight times the size of the rival.
This feature helps AI systems process more information simultaneously. In certain circumstances, the new artificial intelligence requires only about 30% of the resources that DeepSeek would use.
The company used large-scale reinforcement learning to train the M1, employing 512 Nvidia H800 GPUs, with a rental cost of $534,700, according to a statement. She also promised to release more news in the coming days.
MiniMax is supported by large Internet companies in China, such as Tencent and Alibaba Group, and belongs to a select group of about half a dozen domestic AI startups known as the “Little Dragons”. Together, these companies raised billions of dollars in risk financing in the last year. The information is from O Globo
