CoolingModels
Microsoft AI finishes pre-training MAI-Base-2, a 2.8T-parameter MoE model
Microsoft AI said Oct. 15 it finished pre-training MAI-Base-2, a mixture-of-experts model with 2.8T total and 80B active parameters. It claims 4-12x better pre-training FLOPs efficiency than DeepSeek-V4-Pro-Base, on its own held-out evaluations. The model is still in post-training with no release date; Microsoft plans to license weights to trusted partners through Foundry.
- HEAT
- OUTLETS
- 2
- FIRST SEEN
- LAST SIGNAL
Made the five · #5 on Friday, October 16, 2026
Microsoft builds a 2.8T-parameter model of its own
Primary source
Microsoft AImicrosoft.ai/news/frontier-pre-training-at-microsoft-ai/Coverage · 2 articles, oldest first
Thursday, October 15, 2026