中文翻译

摘要: 微软在Build 2026大会上发布首款自研推理模型MAI-Thinking-1等7款内部AI模型,标志其多模型战略正式落地。该模型完全基于商业授权企业数据从零训练,未蒸馏任何第三方模型,有效保障企业数据合规与溯源安全。其采用稀疏混合专家架构,拥有350亿活跃参数与25.6万上下文窗口,在数学推理与编程基准测试中表现优异,性能对标Claude 4系列。目前模型已通过Microsoft Foundr

正文

Microsoft kicked off

Microsoft Build 2026

at Fort Mason Center in San Francisco on Tuesday, June 2, unveiling seven in-house AI models led by MAI-Thinking-1 — the company's first reasoning model, built from scratch on commercially licensed enterprise data with no distillation from third-party models, including OpenAI's GPT series. The keynote, delivered by Satya Nadella beginning at 9:30 a.m. PT, marked the most concrete evidence yet that Microsoft's multi-vendor, multi-model strategy has moved from roadmap to shipping product.

Read more:

GitHub Copilot Replaces GPT-4 With Project Polaris, Ships Multi-Agent VS Code at Build

MAI-Thinking-1: Microsoft's First Reasoning Model Enters Private Preview

MAI-Thinking-1

is a mid-sized model with 35 billion active parameters and approximately one trillion total parameters in a sparse Mixture of Experts architecture, along with a 256,000-token context window — enough, Microsoft said, to process a 600-page document in a single pass. Microsoft trained it entirely on clean, commercially licensed data without distillation from any third-party model, which the company describes as a key enterprise differentiator because it gives customers confidence about the provenance of what the model learned.

model reaches 97.0 percent on AIME 2025 and 94.5 percent on AIME 2026

, benchmarks that test mathematical and multi-step scientific reasoning. On SWE-Bench Pro — a software engineering benchmark — Microsoft says it matches Claude Opus 4.6 on coding tasks, and in blind side-by-side evaluations run by Surge, Microsoft's independent human rating partner, MAI-Thinking-1 was preferred over Claude Sonnet 4.6. Microsoft has published a preprint describing the evaluation methodology, but full reproduction of the results by independent labs has not yet occurred, leaving those benchmark claims open to challenge until confirmed externally.

MAI-Thinking-1 is now available in private preview through Microsoft Foundry. It supports function


来源: Brave/techtimes.com

采集时间: 2026-06-03 20:42:14

AIGPTOpenAI微软人工智能