Microsoft has unveiled Microsoft-Decision-1, a new AI model engineered for rapid, economical structured decision-making. According to the company, the system achieves a median-latency speed roughly 35 times faster than GPT-6 Sol, tailoring it specifically for workflow automation, AI agents, and corporate tools.
In contrast to versatile AI models that craft text or address intricate inquiries, Decision-1 is tailored to pick from predetermined choices and assign probabilities to potential outcomes. This approach is intended to manage standard decisions efficiently, reducing the need to constantly deploy larger, costlier language models.
Microsoft Decision-1: Benchmark Performance and Pricing
Constructed upon the Qwen3.5-9B architecture, Decision-1 received targeted post-training to enhance its decision-scoring capabilities. Data shared by Microsoft indicates that its median-latency performance is also 4.5 times faster than Quyet-1.0-Large.
Microsoft reported that Decision-1 captured the top accuracy ranking in 36 separate benchmarks comprising nearly 150,000 inquiries. These figures, however, originate directly from the company and await independent confirmation.
The system costs USD 0.042 per million input tokens, and output tokens carry no charge. This cost structure is designed to bring down expenses for developers and organizations handling repetitive, high-volume choices.
The Xbox Research division at Microsoft evaluated Decision-1 using upwards of 10,000 user reviews and game feedback items. The evaluation showed that the model matched the quality of GPT-6 Sol while running more than 14 times faster and costing roughly one-two-hundredth as much.
AI Agents and Business Applications
Decision-1 suits tasks like binary choices, multiple-choice queries, ratings, and rubric-based evaluations. Potential use cases encompass content categorization, applicant tracking, task ranking, quality assurance, outcome validation, and safety evaluations.
When integrated into AI agents, Decision-1 can evaluate whether to proceed with a task, repeat a process, escalate a problem, or route it to a human operator. Such functions have the potential to lower operational expenses and streamline turnaround times for high-volume workflows.
Also Read: Oracle vs Microsoft: Comparing Two AI-Driven Cloud Stocks
Security Testing and Availability
Microsoft reported conducting safety evaluations on Decision-1 covering prompt injections, jailbreak vulnerabilities, and toxic content detection. The company additionally measured its output stability when presented with varied inputs, noting a mean decision-change likelihood of 1.3% across eight distinct perturbation tests.
Because these outcomes stem from internal assessments, independent verification is still necessary. Developers can access Decision-1 presently via Microsoft Foundry, with subsequent availability slated for OpenRouter.
This release underscores Microsoft’s strategy to split routine choice selection away from heavy text generation and complex reasoning workloads. Real-world deployments will ultimately determine if the claimed gains in speed, cost-efficiency, and accuracy are sustained outside of controlled testing.
