AI/News
Microsoft launches Microsoft-Decision-1, a model that only decides: a calibrated probability for each of a fixed set of options, for routing and classification, at $0.042 per million input tokens with output free
Microsoft's own 36-benchmark comparison puts it first on accuracy and 35 times quicker than GPT-6 Sol. The chart's footnote says the two latencies were measured in different places, and GPT-6 Sol is "not ranked" on the accuracy axis of the same chart.
By Daily Aletheia · Checked against the primary source · 9 October 2026 · 2 min read
Microsoft introduced Microsoft-Decision-1 on 9 October, "our new model for fast decision-scoring, available in Microsoft Foundry and coming soon through OpenRouter". The post, by Achint Srivastava, VP of Software Engineering in the Office of the CTO, describes a category rather than a chatbot: "decision models are purpose-built to deliver structured outputs that software can immediately act on". It is "designed for routing, classification, prioritization, verification, and workflow control".
What it does: "When given a fixed set of answer options, Microsoft-Decision-1 provides a calibrated probability score for each option." It "supports yes/no, multiple-choice, and rating options, as well as rubric-based grading of AI responses and agent actions". Underneath, "we post trained Qwen3.5-9B for fast, single-pass decision scoring and will soon rebase it on other models, including Microsoft AI (MAI) and OpenAI".
The numbers are Microsoft's. "Microsoft-Decision-1 achieved the highest accuracy in our 36-benchmark comparison, spanning nearly 150,000 questions across benchmarks kept blind from training", and "it was the fastest measured: 4.5 times quicker than Quyet-1.0-Large, the runner-up, and 35 times quicker than GPT-6 Sol". The chart beneath puts the average at 83.5% against Quyet's 81.9%, a median latency of 85 ms against GPT-6 Sol's 3.01 seconds, and a calibration score of 92.2, second to Quyet's 93.1. The chart's own footnote says the other models' latency is "JevBench v1.6.1 adjusted median (p50), checked Oct 7, 2026" while "Microsoft-Decision-1 measured through Foundry in the same region"; GPT-6 Sol sits "not ranked" on the accuracy column. On robustness, the model "changes its decision on 1.3% of perturbations on average with zero flips when option descriptions are paraphrased or when options are reversed or shuffled".
Internal uses named: XBOX Research sorted "more than 10,000 open-ended pieces of feedback and reviews" and found it "competitive on quality with GPT-6 Sol while running over 14 times faster and 200 times less expensive"; the Copilot team found it "competitive with GPT5.6 Luna and 100 times faster". Price: "Input tokens cost $0.042 USD per million tokens. Output tokens are free."