June 3rd message, just nowMicrosoftOrganization of the Build 2026 Conference of Developers, which focused on self-studyModelFAMILY, LOCAL AI DEVELOPMENT EQUIPMENT, NEW AI TERMINAL FORM AND CORPORATE AI GOVERNANCE TOOLS, ANNOUNCING THEIR COMPREHENSIVE LAYOUT AT THE LEVEL OF AI INFRASTRUCTURE。
SEVEN MAI MODELS COVERING REASONING, CODES, IMAGES, VOICE AND TRANSFER DIRECTIONS WERE INTRODUCED AT A ONE-TIME EVENT。
MAI Thinking 1 uses a thin MoE structure, active 35B, total parameter size approximately 1T, supporting 256K token context
The code model MAI Code 1 Flash, trained by Microsoft End-to-end, scored 51.2%, above Claude Haiku 4.5, 35.2%, currently being rolled out to the GitHub Copilot individual users
Image model MAI Image 2.5 has entered PowerPoint and plans to expand to OneDrive
Transcribe 1.5 for 43 languages
The voice model MAI Voice 2 supports 15 languages and can match sound with short samples。
With regard to hardware, Microsoft publishes a development-oriented Surface RTX Spark Dev Box, known by Nadra as "dream Machine"。
Based on the RTX Spark platform, it provides 1 petaflop AI calculus, 20 CPU cores and 128GB unified memory, heat design 100W, and is scheduled to be rolled out this fall. Young Weida CEO Wong In-hoon says via video link that PC is moving from PC to Personal AI。
Microsoft also displays two types of reference equipment codenamed Project Solara:
Desktop working terminals based on the UNDC chip supporting identification, voice interaction and Microsoft 365 Copilot access
The use of high-intensity digital cardboards with chips for mobile work scenes such as medical and field collections。