Microsoft has introduced Decision-1, a new AI model designed for fast, structured decision-making tasks such as classification, evaluation, and agent routing.
Key facts
- •Decision-1 is based on the Qwen3.5-9B language model.
- •The model is designed to guide and control AI agents in complex environments.
- •Microsoft reports an 83.5% accuracy rate and 85 ms latency for the model.
- •Input tokens for Decision-1 cost $0.042 per million, with free output tokens.
- •The model is accessible via Microsoft Foundry and OpenRouter.
Microsoft has entered the AI decision model market with the release of Decision-1. Built on the Qwen3.5-9B architecture, the model is designed to handle classification, evaluation, and routing tasks while guiding agents through complex environments.
By the numbers
Performance and Benchmarking
According to Microsoft, Decision-1 achieved 83.5% accuracy and 85 ms latency across 36 benchmarks involving nearly 150,000 questions. The company claims the model is 2.5 times faster than the H2O-Lightning-4B model. Microsoft's internal testing placed Decision-1 ahead of Jev 1.13.0, though the comparison did not include Cloudflare's open-source Clef models.
Availability and Cost
Decision-1 is currently available through Microsoft Foundry and OpenRouter. The pricing structure for the model is set at $0.042 per million input tokens, while output tokens are provided at no cost.
Advertisement
This article was independently rewritten by ManyPress editorial AI from reporting originally published by The Decoder.
