Microsoft
Microsoft-Decision-1
Launch post: HIPAA-compliant Microsoft-Decision-1: the decision-model layer
Specs
- Input price
- $0.042/M
- Output price
- Free
- Context window
- 32.8K tokens
Microsoft's proprietary 9B decision model - post-trained from Qwen3.5-9B to score a fixed set of questions with calibrated probabilities in a single pass, at a 32,768-token context window.
Capabilities
- Returns a calibrated probability for every option instead of generating text
- Single-pass scoring over a shared prefix; output tokens are free
- 32,768-token context and typed questions: choice, yes or no, and ordered rating
- 1.3% average decision-flip rate across eight perturbation types in Microsoft's evaluation
- Runs in public preview on Microsoft Foundry, served through our BAA program
Best for
- Input triage or queue routing before a generative model is called
- Agent guardrails and human-review thresholds
- Quality-review rubrics and coding support where a confidence score matters
Limitations to keep in mind
- Proprietary and not open weights - cannot be self-hosted
- 32K context is a fragment, not a full chart
- Returns a distribution, not a rationale
- Calibration is dataset-dependent - plot a reliability diagram on your own cases
- Preview lifecycle: the question schema, pricing, and model identifier can change
HIPAA-compliant hosting
Microsoft-Decision-1 is available under our HIPAA-compliant design-partner program, with a signed Business Associate Agreement (BAA), encryption in transit and at rest, and access controls. We are currently onboarding design partners, with general availability coming soon. Run the model behind a unified OpenRouter-style API and swap to another model with a single parameter. For the deployment requirements behind that, see our 2026 guide to HIPAA compliance for AI & LLM APIs.
Design partner program
Run Microsoft-Decision-1 under HIPAA
Join the waitlist to be prioritized. We'll reach out with a qualification call and early access.