AIToday
Large Language ModelsAI Coding AssistantsSiliconANGLE AIPublished: Nov 18, 2026, 06:01 JST

Microsoft ships Decision-1.1 and new MAI models for Foundry

Send on LINE
Microsoft ships Decision-1.1 and new MAI models for Foundry

3 Key Points

  1. What happened

    Microsoft announced general availability of Decision-1.1, a decision model built on Google's Gemma 4 26B A4 base with multimodal image support, plus new MAI models including MAI-Code-2 and MAI-Code-2-Flash, now on GitHub.

  2. Why it matters

    Decision-1.1 is positioned to make software decisions dynamically at low cost rather than by 'thinking it through' like large language models, and provides calibrated confidence scores so applications can defer when uncertainty is high, which could lower decision costs for enterprises.

WHO IT HITSEnterprise developers and IT teams building AI agents on Microsoft Foundry, who can now choose between performance and cost with Decision-1.1 and the MAI family, likely affecting how they route, classify and verify tasks at scale.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

Microsoft's rollout builds on the previous Microsoft-Decision-1, released in October, and positions Decision-1.1 as a significant improvement in intelligence and speed at a competitive price. The shift to a Gemma 4 26B A4 base adds multimodal support, meaning the model can analyze images in instructions, which may broaden its use beyond text-only routing and classification.

The Foundry updates extend beyond the decision model. Human annotations, now in preview, allow experts to rate and comment on agent traces based on business judgment, and this feeds back into an improvement loop combined with optimization tests. Foundry also gains capabilities to adapt supported models with organizational data using reinforcement learning while it runs compute, with each improvement becoming the next baseline.

The expanded model pool, featuring the top 50 models and more than 11,000 models from OpenAI, Anthropic, Meta, Mistral AI and others, suggests Microsoft is positioning Foundry as a neutral ground for enterprises to evaluate and deploy models from multiple providers under one continuous improvement workflow.

FAQ
What is Decision-1.1 built on?
It is built on Google LLC's Gemma 4 26B A4 base and adds multimodal support so it can analyze images in instructions.
Where are the new MAI models available?
MAI-Code-2 and MAI-Code-2-Flash are generally available on GitHub, with Code-2-Flash in private preview in Foundry; MAI-Image-2.7 and MAI-Image-2.7-Flash are in preview.
How does the continuous improvement loop work for agents?
Foundry uses internal evaluations to set baselines for quality, safety, latency and cost, and human annotations in preview let experts rate agent traces, feeding back into an improvement loop.
SiliconANGLE AIRead Original Article
Send on LINE

Also reported by Publickey

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleJensen Huang exit seen shifting HBM demand beyond Nvidia