Microsoft completes pre-training of 2.8T-parameter MAI-Base-2 model

Microsoft has finished pre-training MAI-Base-2, a mixture-of-experts foundation model with 2.8 trillion total and 80 billion active parameters. Training ran for 38 days on 16,000 Nvidia GB200 GPUs over 50 trillion tokens, and Microsoft estimates its approach is 4.2 to 12.3 times more compute-efficient than DeepSeek-V4-Pro-Base at matching performance.

• 2.8T total parameters, 80B active per input
• 16,000 Nvidia GB200 GPUs over 38 days, 50T tokens
• Images make up 18% of training mix; 36% more unique data than MAI-Base-1
• Weights to be licensed to partners via Microsoft Foundry, not released publicly