Microsoft Unveils Aion 1.0 On-Device AI Models at Build 2026

Microsoft Brings AI Home with Aion 1.0

Microsoft opened its Build 2026 developer conference on Tuesday with a headline announcement: Aion 1.0, a new family of on-device AI models built to run locally on Windows PCs. The launch marks a significant step in the company's strategy to embed intelligence directly into the operating system, reducing the reliance on cloud-based processing that has defined the current generation of AI assistants.

Two Models, Two Jobs

CEO Satya Nadella introduced two members of the Aion 1.0 lineup during the opening keynote. The first, Aion 1.0 Instruct, is a small language model that replaces the previous Phi-4-mini inside Microsoft Edge. Its smaller, faster footprint extends compatibility to devices with less capable GPUs, and even to machines running purely on CPU inference. A developer preview is already available in the Edge Canary and Dev channels, with an open-source release planned for Hugging Face in July.

The second model, Aion 1.0 Plan, is a far larger 14-billion-parameter reasoning and tool-calling model aimed squarely at agentic workflows. Built into Windows itself, it lets applications reason over user intent, invoke tools, manage files, and coordinate sub-agents directly on local hardware. Microsoft says the model will arrive in the coming months.

Windows AI Across Every Chip

Alongside the models, Microsoft announced an expansion of its Windows AI APIs to run across CPUs, GPUs, and NPUs from all four major silicon vendors: AMD, Intel, Qualcomm, and Nvidia. The move allows developers to tap local compute rather than defaulting to cloud inference, improving privacy, cutting latency, and eliminating the per-query costs that come with cloud APIs. Nadella framed the initiative as taking AI use and development away from the cloud and onto one's PC.

New Hardware for Developers

The conference also introduced the Surface RTX Spark Dev Box, a compact developer workstation powered by Nvidia's RTX Spark silicon. The device delivers up to 1 petaflop of AI performance with 128 gigabytes of unified memory, and is capable of running models with up to 120 billion parameters locally, all within a 100-watt power envelope.

A Broader Push

Aion 1.0 sits within a wider wave of Build 2026 announcements, including Microsoft Discovery for scientific R&D and the company's first in-house reasoning model, MAI-Thinking-1. Build 2026 runs 2 to 3 June at the Fort Mason Center in San Francisco, with sessions covering agentic AI, Azure AI Foundry, GitHub Copilot, and responsible AI tooling. Together, the announcements signal a clear direction: powerful AI that lives on your device, not just in a distant data centre.