Microsoft

Microsoft builds 7 in-house "MAI" models at Build 2026: less reliance on OpenAI, Copilot becomes a "super app"

A developer working across multiple code screens

At Build 2026 (which opened on 02/06 in San Francisco), Microsoft introduced 7 self-developed models called "MAI" for coding, reasoning, image and voice tasks — according to reports, these include MAI-Thinking-1 (reasoning) and MAI-Image-2.5 (already running in PowerPoint). Microsoft is also consolidating its AI assistants into a "Copilot super app".

Quick summary

  • When: Build 2026, opened on 02/06/2026.
  • What: 7 self-developed 'MAI' models (coding, reasoning, image, voice) — according to reports.
  • Highlights: MAI-Thinking-1 (reasoning) and MAI-Image-2.5 (in PowerPoint).
  • Why it matters: even Microsoft is building its own models to reduce single-vendor dependence.

What happened?

According to reports from Build 2026, Microsoft is doubling down on its in-house 'MAI' model line alongside its relationship with OpenAI. Building its own models for many tasks and folding the assistant experience into a single 'Copilot super app' signals a strategy of being more proactive about core AI capabilities. (Detailed specs are based on reports and are for reference only.)

A robotic hand touching a digital network — abstract AI
Microsoft is doubling down on its in-house model line alongside OpenAI. Photo: Tara Winstead / Pexels

Why this matters

When one of the largest tech companies also builds its own models to reduce reliance on a single vendor, that is a strong signal of the value of owning your AI capabilities. Depending entirely on one party is a strategic risk.

A software engineer writing code at a desk
Building your own models to reduce reliance on a single vendor. Photo: mikhail-nilov / Pexels

A view for businesses

Vietnamese businesses don't need to train models from scratch like Microsoft, but they can achieve a similar goal with in-house AI built on open-source models running on-premise — owning the infrastructure and data, and avoiding lock-in to any single vendor.

Frequently asked questions

What is MAI?

It is a line of models developed in-house by Microsoft, introduced/highlighted at Build 2026 for many tasks such as coding, reasoning, image and voice. Detailed specs are according to reports.

Is Microsoft dropping OpenAI?

Not exactly. This is a strategy to add in-house model capabilities to reduce dependence on a single vendor, in parallel with the existing relationship.

What can small businesses learn?

Diversifying and owning your AI capabilities helps reduce vendor lock-in risk. In-house AI built on open-source models is a feasible way to achieve that.

Own your AI

Namtech deploys in-house AI on open-source models — with no lock-in to any single vendor.

Book a free consultation

Note: This article is compiled from public sources as of 23/06/2026; information is for reference and may change.

Get started

Start with a free assessment

To define the right package and detailed scope, Namtech offers a short, no-cost assessment.

We reply within 1 business day. No spam, we never share your info.