Summary
MAI-Base-2 contains 2.8 trillion total parameters and activates 80 billion per token. Microsoft says it outperformed the open-weight models tested on held-out evaluations and achieved substantially better pre-training compute efficiency, but the model remains in post-training and is not yet generally available.
Why it matters
The announcement shows Microsoft building a large in-house model pipeline at frontier scale while treating training efficiency, data provenance and future Foundry licensing as strategic assets.