Microsoft has completed pretraining MAI-Base-2, a new AI model with 2.8 trillion total parameters and 80 billion active parameters. They claim it outperformed every open-source model evaluated while requiring up to 12 times less computing power to match a leading competitor. Built on a mixture-of-experts architecture, the model activates only a portion of its parameters for each task. Microsoft said its training approach was 4.2–12.3 times more compute-efficient than DeepSeek-V4-Pro-Base across four benchmarks and 6.3–12.7 times more efficient than its predecessor, MAI-Base-1. The model was trained using 16,000 Nvidia GB200 GPUs over 38 days in August and September, processing approximately 50 trillion training tokens. Its unique-token dataset increased by around 36% compared with MAI-Base-1, released in June, while image-related material accounted for 18% of the training mix following the integration of native vision capabilities. $Microsoft(MSFT)$
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
