[CLSTR]
< Back to situations

Monitor this situation.

[SITUATION] · [ACTIVE] · [TECHNOLOGY]

2 clusters · 7 sources · 3 days · First seen · Last updated

Microsoft MAI-Base-2 model development

Overview

Microsoft has announced the development of MAI-Base-2, a next-generation foundation model utilizing a Mixture-of-Experts (MoE) architecture. Initial reports indicated the model features 2 trillion parameters and was trained over 38 days using 16,000 Nvidia GB200 chips. The model functions as a vision-language model designed for tasks such as spatial reasoning and scientific chart analysis.

Subsequent details revealed the model possesses 2.8 trillion total parameters. Microsoft AI CEO Mustafa Suleyman stated that MAI-Base-2 outperforms OpenAI’s GPT-5.6 Sol in pattern recognition and information prediction while requiring only one-quarter of the computing power used by OpenAI’s equivalent models. This development is viewed as an effort to reduce Microsoft’s reliance on OpenAI.

Alongside the model launch, Microsoft Research introduced StudentSim, a system designed to evaluate AI tutoring tools by using AI-driven student simulators to measure learning outcomes rather than simple output accuracy.

Entities

Microsoft · MAI-Base-2 · Mustafa Suleyman · StudentSim · DeepSeek

Timeline

  1. [TECHNOLOGY] 2 sources
    Microsoft unveils MAI-Base-2 AI model and StudentSim educational research

    Microsoft has unveiled MAI-Base-2, an efficient new AI model outperforming OpenAI's GPT-5.6 Sol, and StudentSim, a research system that uses AI student simulators to improve the effectiveness of AI tutoring.

  2. [TECHNOLOGY] 5 sources
    Microsoft unveils MAI-Base-2 foundation model with 2T parameters

    Microsoft has unveiled MAI-Base-2, a 2-trillion parameter foundation model using MoE architecture to achieve up to 12x greater hardware efficiency than leading open-source alternatives.

Sources

drweb.de · game.techbang.com.tw · intelligentliving.co · prodiris.fr · siliconangle.com · unwire.pro · zdnet.co.kr