Microsoft released Microsoft-Decision-1 with callable Foundry and OpenRouter endpoints and public pricing.
Microsoft releases a fast model for structured decisions
2026-10-09Verified announcementTRIAL
Microsoft released Microsoft-Decision-1 for routing, classification, ranking, verification and workflow control over fixed choices with calibrated probabilities. The post-trained Qwen3.5-9B model is available through Microsoft Foundry and OpenRouter. Microsoft reports the highest accuracy across 36 blind benchmarks, roughly 35 times lower P50 latency than GPT-6 Sol and input pricing of $0.042 per million tokens.
Availability: Microsoft Foundry and OpenRouter
What this means for you
For agent systems that make many binary, routing or guardrail decisions, a specialized decision model could be cheaper and more stable than a general LLM. Core results remain vendor-reported, and closed-choice scoring cannot replace open-ended reasoning; validate accuracy, calibration, latency and false refusals on real routing data.
Evidence & related material
Technical details & scores
| Base Model | post-trained Qwen3.5-9B |
|---|---|
| Interface | fixed choices with a calibrated probability for each option |
| Availability | Microsoft Foundry and OpenRouter |
| Pricing | $0.042 per million input tokens; output tokens free |
| Vendor Evaluation | 36 benchmarks and nearly 150,000 questions kept blind from training |
| Vendor Latency | about 35x faster P50 than GPT-6 Sol |
| Robustness | 1.3% average decision flips across eight perturbations |
| Safety Evaluation | 5,250 requests across 11 safety benchmarks |
Technical impact 4
Engineering value 5
Adoption signal 2
Maturity 4
Verification cost 2
Risk 3