$2 / month

Comment and reply across TML.

Start or manage membership

Most pulsed

Latest from other desks

tml.the modern leap
Tech / AI

Microsoft Ships Specialist Models For Coding And Security

At Ignite in San Francisco, Microsoft made its MAI Code 2 and Code 2 Flash models generally available in GitHub Copilot and previewed Cyber 2 Flash, a security model it says beats frontier models at a fraction of the cost per task.

TML Tech DeskNov 18, 4:08 PM ET · 2 min read

Edited and fact-checked by Stephen Sprague

Microsoft chief executive Satya Nadella speaking at a conference (archive). Photo: OFFICIAL LEWEB PHOTOS, via Wikimedia Commons (CC BY).
Summary

The models are live in Copilot today, and the first honest benchmark will come from the repositories that run them this week, not from the keynote.

Tap a passage. React to the part that matters.

Satya Nadella opened Microsoft's Ignite conference in San Francisco on Tuesday, November 17, with a keynote built around smaller, cheaper, specialized models. The headline products are not bigger general models at all. They are narrower ones, built for coding, security and routine decisions, and priced to run constantly inside enterprise software.

MAI Code 2 is the flagship of the coding family, and Microsoft says it handles the work of taking code to production, deep debugging, multifile implementation and repo-wide reasoning. It is generally available today in GitHub Copilot, both in the model picker and inside Copilot Fusion, and in Microsoft Foundry. Code 2 Flash covers the fast work, the code explanations, targeted edits and routine fixes, at a lower cost.

Every performance figure in the announcement is Microsoft's own, and that is the part of the launch worth reading twice. The company says Code 2 sits at the frontier of cost-performance for everyday enterprise coding, and that Cyber 2 Flash performs ahead of frontier models at a fraction of the cost per task, but it published no independent benchmark in the keynote materials. Buyers get the claim and the invoice, and the verification is left to their own test suites.

Cyber 2 Flash is the security entry, a flash-sized model Microsoft calls its most capable security model yet, and it is being co-designed with Perception, the company's agentic security system. Perception runs teams of red, blue and green agents through playbooks, attacking, detecting and remediating at machine speed while a human stays in control of the outcome.

Agent 365 is where the money controls live, because Microsoft says more than 50 million agents are already registered in the control plane that extends existing identity, governance and security systems to agents. The new Business Analytics layer ties agent spending to business outcomes, and Microsoft describes it as an ERP system for AI usage, while Cost Management adds the spending policies, alerts and forecasting that administrators asked for.

Maia 300, Microsoft's own AI chip, is the hardware underneath the pitch, and the company says it will deliver 40 percent more tokens per watt than leading GPU-based systems. Microsoft also announced MAI Licensed Weights, which gives trusted customers access to Microsoft AI model weights through its Frontier Co.

For a working developer, the change that lands today is the model picker, where Code 2 and Code 2 Flash are live in Copilot now, priced below the frontier models they sit beside. Every repository they touch will test Microsoft's cost claims against real code within the week.

Sources

The conversation

A little more perspective.

Reading is free. Verified members can comment and reply for $2 a month.

The conversation.

Loading comments…

Keep your curiosity going.

Some links here are affiliate links. If you buy through one, we might make a commission — it never changes what you pay. As an Amazon Associate I earn from qualifying purchases.