# Microsoft launches Decision-1, a model that scores fixed options for agents

published: 2026-10-12T13:41:53.211Z
topic: infrastructure
source: https://www.producthunt.com/products/microsoft-decision-1
url: https://aiagentnews.io/stories/microsoft-launches-decision-1-a-model-that-scores-fixed-options-for-agents

Microsoft has made Decision-1 available in Microsoft Foundry, a model that returns a calibrated probability for each of a fixed set of options in a single pass. It targets routing, classification, verification and agent control rather than text generation.

Decision-1 is a decision-scoring model post-trained from Qwen3.5-9B. Given fixed options, it returns structured output that software can act on directly, without any text to parse. Microsoft says it is about 35 times faster than GPT-6 Sol at median latency, and that in its tests it was 4.5 times faster than the runner-up, Quyet-1.0-Large.

Microsoft reports the highest accuracy in a 36-benchmark comparison covering nearly 150,000 questions kept blind from training. It also says the model changed its decision on 1.3% of cases on average across eight types of input changes, and never when option descriptions were paraphrased or options were reversed or shuffled. Safety testing covered 5,250 requests across 11 benchmarks.

Microsoft says the Xbox Research team, the Copilot team and Microsoft Discovery already use the model. Pricing is $0.042 per million input tokens with output tokens free, and OpenRouter availability is listed as coming soon.

## Why it matters if you build agents

If your agent routes tasks or gates actions with an LLM call, a single-pass scorer could reduce latency and cost on those steps. Treat the probabilities as measurements to verify, not fixed guarantees: a commenter on the listing noted that calibration only holds against the data a model was tested on. The listing does not describe a way to monitor calibration drift, so check calibration on your own labeled traffic before acting on the numbers directly.