Skip to content
AI HOT (Curated Pool)

Microsoft releases MAI-Thinking-1 model

微软发布 MAI-Thinking-1 模型

Microsoft released MAI-Thinking-1, an MoE model with 35B active parameters and 1T total parameters, pretrained from scratch on 30T tokens without third-party model distillation.

Why it matters: Microsoft's MAI-Thinking-1 carries 1 trillion total parameters and 35 billion active per pass, pretrained from scratch on 30T tokens with no third-party distillation, making its MoE sparsity and in-house training path the draw.

Read the original ↗Export Markdown