Microsoft AI ·

Microsoft unveils MAI-Base-2, a MoE model with 2.8T total and 80B active parameters, claiming 4x-12x more pretraining FLOPs-efficiency than DeepSeek-V4-Pro-Base

In June, we announced MAI-Thinking-1, a general-purpose foundation model post-trained on top of MAI-Base-1 with strong reasoning performance.

In this story DeepSeek-V4-Pro-Base MAI-Base-2 Microsoft

Microsoft unveils MAI-Base-2, a MoE model with 2.8T total and 80B active parameters, claiming 4x-12x more pretraining FLOPs-efficiency than DeepSeek-V4-Pro-Base

Lead Source

Microsoft AI — Microsoft unveils MAI-Base-2, a MoE model with 2.8T total and 80B active parameters, claiming 4x-12x more pretraining FLOPs-efficiency than DeepSeek-V4-Pro-Base ↗

How this story grew

Coverage · 0 Discussion · 0

Oct 16Oct 17

More

Business Insider: Business Insider

Constellation Research: Constellation Research

Yahoo Finance: Yahoo Finance

Discussion

42 points, 4 comments

Suqiang Song Ashley Stewart

@mustafasuleyman @hanxiao_6 @bentheegg @chrisgpt @scaling01 @choblin29 @mahendrakariya @mustafasuleyman

Related stories

Command Line: Microsoft unveils Microsoft-Decision-1, a fast decision-scoring model trained on Qwen3.5-9B, and says it will soon rebase it on MAI, OpenAI, and other models

Microsoft AI: Microsoft launches MAI-Transcribe-2-Streaming, a model for low-latency, real-time transcripts, and two new voice models, MAI-Voice-2.1 and MAI-Voice-2.1-Flash

Reuters: DeepSeek debuts DeepSeek-V4.1-Flash, its smallest model built on its new Causal Encoder-Decoder architecture, with 552B backbone parameters and 1M-token context

Bloomberg: Tencent releases Hy4 Preview, a 770B-parameter open model with 1M context window, and says it outperforms Z.AI and Moonshot models in internal tests

The Information: Nvidia is developing a Nemotron 4 model with 1T+ parameters, up from Nemotron 3 Ultra's 550B parameters but smaller than leading Chinese open models

Discuss with AI

ChatGPT ↗ Claude ↗

← Back to TechSnif