Microsoft-Decision-1.1 Goes Generally Available, a Small Model Built to Route Agent Tasks

Microsoft announced on 17 November 2026 that Microsoft-Decision-1.1 is generally available. The model is built on Gemma 4 26B A4B, adds image input and handles routing, classification and workflow control.

Christopher Ort
By Christopher Ort · 2 min read
Share
Railway tracks branching in different directions at dawn.
Image: AI-generated, i10X
On this page
  1. Quick Take
  2. In plain English
  3. What happened
  4. Background
  5. How it works
  6. Why it matters
  7. What's still open
  8. Sources
  9. i10x Perspective

Microsoft-Decision-1.1 is generally available in Microsoft Foundry, OpenRouter and Vercel, Microsoft company page said at Ignite on 17 November 2026. Satya Nadella announced it in his keynote as a model for fast, low-cost decisions inside software.

i10X Quick Take

Quick Take

  1. What: a model for structured decisions such as routing, classification, prioritization, verification and workflow control.
  2. Base model: Gemma 4 26B A4B, with new multimodal support for image inputs.
  3. Score: 68 on the Artificial Analysis Decision Index v2.1, up six points from the first version, according to Microsoft.
  4. Speed: more than twice as fast as Microsoft-Decision-1, at the same price.
  5. Use: more than 4,000 customers have already used it, Nadella said.

In plain English

Not every AI task needs a giant model. Many agent steps are small choices, such as which tool to call or which team should get a request. This model is built only for those choices and reports how confident it is, so an app can hand hard cases to a bigger model. It is meant to be cheap and quick enough to run constantly.

What happened

Nadella said: “We have added multimodal support. And we have improved performance, making decisions more accurate and responses faster.” He added that Microsoft uses it internally “for core tasks like routing, data labeling, personalization, content filtering and hundreds of other use cases.” In Copilot it routes requests to Copilot or Cowork, and in Cowork it powers Auto, which adjusts reasoning effort per task.

Background

Microsoft-Decision-1 was released in October. The Book of News says the new version scores ahead of Jev 1.14, Google’s Gemini-Decision-1 and OpenAI’s GPT-6 Luna (Decisions) on that index. It sits alongside Laya and a Decisions API in Foundry.

How it works

The model chooses among permitted actions within developer-defined constraints and returns calibrated confidence scores, so applications can defer when uncertainty is high. Frontier models can then be reserved for complex reasoning.

Why it matters

For developers, a cheap decision layer cuts the cost of agent loops.

For Microsoft, it makes Copilot’s model routing a visible product feature.

For rivals, Google and OpenAI now have named competitors in a decision-model niche.

What’s still open

  • Independent verification of the benchmark score, which Microsoft cites.
  • Per-token pricing, which Microsoft calls competitive without listing it here.
  • How well it handles edge cases outside routine routing.

Sources

  • Microsoft, “Ignite 2026: Building the frontier ecosystem,” Satya Nadella, The Official Microsoft Blog, 17 November 2026
  • Microsoft, “Microsoft Ignite 2026 Book of News,” 17 November 2026

i10x Perspective

The benchmark comparison is Microsoft’s own summary. The more interesting point is architectural: small specialist models are being placed in front of large ones to control cost. More in our Models section and Agents section.

Christopher Ort

Christopher Ort

Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU). He is the author of “The Style Penalty”, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.

Timeline

What happened before

  1. Release

    MAI-Code-1.1-Flash

    256K context

    microsoft.ai (source, opens in a new tab)
  2. Partnership

    Microsoft commits multiple billions to Mistral’s European GPU capacity

    With Microsoft · Compute

    reuters.com (source, opens in a new tab)
  3. Release

    MAI-Thinking-1 (7-model MAI family)

    Flagship reasoning model plus MAI-Code-1-Flash, MAI-Image-2.5, MAI-Transcribe-1.5 and MAI-Voice-2.

    microsoft.ai (source, opens in a new tab)
  4. Partnership

    Microsoft and OpenAI amend deal, ending exclusive licence

    With OpenAI · Cloud

    openai.com (source, opens in a new tab)
  5. Partnership

    Microsoft and NVIDIA to invest up to $15B in Anthropic

    With NVIDIA, Anthropic · Compute · $15B

    Microsoft (up to $5B) and NVIDIA (up to $10B) to invest in Anthropic; Anthropic commits to $30B of Azure compute.

    anthropic.com (source, opens in a new tab)
Full Microsoft timeline ›

Editor's picks

  1. Microsoft Teams Lets AI Agents Join as Digital Teammates With Their Own Identity
  2. Azure Cobalt 200 VMs Reach General Availability With Up to 50% More CPU Performance
  3. Microsoft Launches AI Transparency Initiative With NIST Standards Work and 40 Partners
  4. Agent 365 Gets Business Analytics and FinOps for AI as Nadella Cites 50 Million Agents
  5. Microsoft Perception Is Generally Available, With Red, Blue and Green Security Agents