Microsoft launches Decision-1 to compete in fast-growing AI decision model market

Microsoft joins the decision model race

Microsoft has entered the rapidly expanding field of AI decision models with the launch of Decision-1, a system built for quick, structured choices like classification and routing. Based on Alibaba's Qwen3.5-9B language model, Decision-1 claims top accuracy and speed in internal benchmarks ³. The release highlights a growing trend: specialized AI tools that make fast decisions rather than generate lengthy text.

What makes a decision model?

Decision models are a new AI category that score fixed answer options and return probabilities, letting software act immediately without parsing generated text. Unlike large language models (LLMs) that write prose, decision models output structured data like JSON, making them ideal for routing user queries, verifying content, or controlling AI agents. Microsoft frames Decision-1 as purpose-built for these decision-scoring roles ².

Performance and benchmarks

In Microsoft's testing across 36 benchmarks covering nearly 150,000 questions, Decision-1 achieved the highest average accuracy at 83.5%, beating other decision models and LLMs ³. It also recorded a p50 latency of 85 milliseconds, making it 2.5 times faster than runner-up H2O-Lightning-4B and 35 times faster than GPT-6 Sol ². However, some benchmarks were vendor-run, and calibration scores varied, with Decision-1 placing second to Quyet-1.0-Large .

Pricing and availability

Decision-1 is available through OpenRouter and Microsoft Foundry, priced at $0.042 per million input tokens with output tokens free. This low cost aims to make decision intelligence accessible for developers integrating AI into workflows. Microsoft plans to rebase future versions on other models like MAI and OpenAI's systems ³.

Competitors and context

The decision model trend started with startup Jev in mid-September, prompting rapid adaptation from open-source communities. Microsoft's entry follows competitors like OpenAI and Cloudflare, which have also released decision-focused models. Satya Nadella announced the launch on X, stating the model is already being tested across Microsoft for incident management and other tasks ¹.

Key facts

• Decision-1 is based on Alibaba's Qwen3.5-9B language model
• Achieved 83.5% average accuracy across 36 benchmarks
• 2.5 times faster than runner-up H2O-Lightning-4B
• Priced at $0.042 per million input tokens with free output tokens
• Available via Microsoft Foundry and OpenRouter

Seen on social

Introducing Microsoft-Decision-1, our new model for fast decision-making. It delivers top performance on structured decision tasks, outperforming both LLMs and other decision models in latency and quality. We’re already testing it across Microsoft for everything from incident… pic.twitter.com/M3fAOD90r6

— Satya Nadella on X, see the post

Sources

1. Microsoft's Decision-1 model enters the fast-growing AI decision model race — the-decoder.com
2. Microsoft-Decision-1: Our model for fast decision-making
3. Microsoft AI Releases Microsoft-Decision-1: A Qwen3.5-9B Decision ... — marktechpost.com
4. Introducing Microsoft-Decision-1 in Microsoft Foundry for decision and ...
5. Microsoft Decision-1 in Foundry: Fast AI Classification Model ... — windowsforum.com