2.8 TRILLION PARAMETERS, 80 BILLION ACTIVE — 12X DEEPSEEK'S EFFICIENCY

Microsoft AI announced it has finished pre-training MAI-Base-2, a mixture-of-experts foundation model with 2.8 trillion total parameters and 80 billion active parameters. The company says it outperforms every open-weight model it tested on held-out evaluations and is 4 to 12 times more pre-training FLOPs-efficient than DeepSeek-V4-Pro-Base, the strongest open base model it compared against. The model was trained on 16,000 GB200 GPUs over 38 days in August and September 2026.

Microsoft says the model was trained on organic data without distillation from other models, giving it clear data provenance, and that it plans to license the weights to trusted partners through Foundry. MAI-Base-2 is now in post-training. The company attributes the efficiency gains to scaling ladders that account for repeated data, joint tuning of model and infrastructure, improved mixture-of-experts and Muon implementations, and expanded data pipelines including vision data.