Microsoft-Decision-1
Microsoft's Qwen3.5-9B-based decision-scoring model returns calibrated option probabilities in one pass, at $0.042 per million input tokens.
Overview
Microsoft released Microsoft-Decision-1 on October 9, 2026, a model post-trained from Qwen3.5-9B for single-pass decision scoring: given fixed options, it returns a probability for each rather than generating text. It targets routing, classification, prioritization, verification and rubric grading of AI responses and agent actions. Microsoft reports the highest accuracy in a 36-benchmark comparison (about 150,000 held-out questions) and 2.5x the speed of the runner-up, H2O-Lightning-4B v1.1.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- Returns calibrated probabilities for yes/no, multiple-choice and rating options
- Post-trained from Qwen3.5-9B; Microsoft plans to rebase it on MAI and OpenAI models
- Changes its decision on 1.3% of perturbed inputs, and on none when options are reordered or paraphrased
- Output tokens are free; only input is billed
- Available in Microsoft Foundry and OpenRouter
Use cases
- Routing and classification steps in agent workflows
- Rubric-based grading of LLM responses and agent actions
- Prioritization and verification gates in enterprise pipelines
Pricing
$0.042 per million input tokens; output tokens free.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 9 Oct 2026
- Record updated
- 15 Oct 2026
- Record last checked
- 15 Oct 2026