Microsoft reveals 3T-parameter MAI foundation model
Microsoft has fully trained its in-house 3T-parameter MAI-Base foundation model, with MAI-Base-2 pre-trained on 16,000 Nvidia GB200 GPUs for 38 days, aiming to reduce dependence on external models.
2026-10-18 ~ 2026-10-19 · 2 related posts
- Episode 1: Microsoft AI completes pre-training of homegrown 2.8T-parameter MoE base model MAI-Base-2(2026-10-16, 12 posts)
- Episode 2: Reported Microsoft MAI-Base-2: 2.8T-Parameter MoE Trained on 16,000 GB200s(2026-10-16, 3 posts)
- Episode 3: Microsoft reveals 3T-parameter MAI foundation model(2026-10-18, 2 posts)
- Microsoft finishes pre-training 2.8T-param MAI-Base-2 on 16,000 GB200 GPUs in 38 days — emmanuelvivier · 2026-10-18
- Microsoft MAI reveals in-house 3T-parameter base models with native encoder-free VLM pretraining — burny_tech · 2026-10-19