Microsoft unveils MAI-Base-2: 2.8T-param MoE, 4-12x more pre-training FLOP-efficient than DeepSeek
ChengleiSi · x · 2026-10-16
Microsoft AI's official blog Frontier Pre-training at Microsoft AI reveals its newest foundation model, MAI-Base-2: an MoE base model with 80B active / 2.8T total parameters, whose pre-training is now complete.
Key facts:
- MAI-Base-2 outperforms every open-weight model tested on Microsoft's held-out evaluations
- It is 4-12x more pre-training FLOPs-efficient than the strongest open base model, DeepSeek-V4-Pro-Base; the savings were reinvested into training a larger model
- Back in June, Microsoft announced MAI-Thinking-1, a general-purpose reasoning model post-trained on MAI-Base-1
The post includes charts comparing MAI-Base-2 against other trillion-plus-parameter models across pre-training benchmarks.
Related event: Microsoft Unveils MAI-Base-2, 2.8T-Parameter MoE Base Model(3 posts)→
More from Infra
- Meta's 816B open-weight Muse Spark Open hits Groq preview at 350+ tokens/sec — GroqLLC · 2026-10-16
- Together AI serves Muse Spark Open serverless for agentic coding and multimodal workloads — togethercompute · 2026-10-16
- Musk says SpaceX-Nvidia 'Star Mind' space datacenter computer hits volume production next year — elonmusk · 2026-10-16
- Muse Spark Open lands on MLX just 30 minutes after release — pcuenq · 2026-10-16
- TSMC earnings note cross-referenced with WFE model and power checks in CS Atlas — BenBajarin · 2026-10-16
- Vercel Drives is GA: persistent storage lets stateful agents survive environment changes — evilrabbit_ · 2026-10-16