» Unfair Advantage
Daily Brief

Microsoft ships its own AI models, and undercuts OpenAI on price

2026-06-05 · Unfair Advantage Editorial

At its Build conference, Microsoft stopped reselling other people's AI and started shipping its own. The MAI family spans coding, reasoning, voice and image, and runs on Microsoft's Azure data centres, which lets it cut the price. MAI-Code-1-Flash is live now in GitHub Copilot and VS Code. The reasoning model, MAI-Thinking-1, is in private preview through Microsoft Foundry. The headline claim, from Microsoft AI chief Mustafa Suleyman: after tuning for McKinsey's tasks, MAI beat OpenAI's GPT-5.5 on quality at ten times better cost efficiency.

Why it matters

Microsoft owns the tools small teams already pay for: Copilot, VS Code, Microsoft 365. If its in-house models are genuinely cheaper to run, that pressure shows up in your bill, and it gives you a real alternative to being locked into one frontier lab. The catch is that a model tuned to run cheap on Azure is still lock-in, just a different vendor's.

Network impact

LatencyMAI-Code-1-Flash is built for low-latency inference and runs in Microsoft's own data centres, which should keep coding responses fast inside Copilot and VS Code.
SecurityFoundry lets you fine-tune MAI-Thinking-1 on your own data; treat that data path with the same care as any other model vendor before sending sensitive material.
ScalabilityRunning the models on Azure rather than paying a third party gives Microsoft headroom to scale capacity and pass lower token costs to developers.

What to do

  1. If you use GitHub Copilot or VS Code, switch the model to MAI-Code-1-Flash on a real task and compare output quality side by side.
  2. Track your token spend before and after the swap. The whole pitch is cost, so measure it.
  3. Don't rip out OpenAI or Claude yet. Run MAI alongside for a week and keep what wins.
  4. If you build apps, request MAI-Thinking-1 access in Microsoft Foundry to test the reasoning model on your own data.
  5. Remember the lock-in risk: cheap models that only run well on Azure tie you to one cloud.

Sources

« All articles