Microsoft / MarkTechPost / The Register
Important
MicrosoftDecision ModelsPricingAgentsMicrosoft ships Decision-1 on Qwen3.5-9B at $0.042/M input, free output
October 9, 20262 min read
Microsoft released Microsoft-Decision-1, a decision-scoring model post-trained from Alibaba’s Qwen3.5-9B for routing, classification, and workflow control. It returns calibrated probabilities for fixed options in a single pass — priced at about $0.042 per million input tokens with free output, matching TypeSafe’s Jev, and claims top accuracy across 36 blind benchmarks at ~35x the speed of GPT-6 Sol.
Why it matters
The cheap decision layer is now a three-way race: TypeSafe Jev, OpenAI Decisions API, and Microsoft’s Qwen-based entry. Agent stacks will route most micro-choices away from frontier models; Microsoft is willing to ship on Chinese open weights first and rebase later.
Microsoft announced Microsoft-Decision-1 on October 9 for public preview in Microsoft Foundry (and on OpenRouter): a decision-scoring model that takes a situation plus a fixed set of answer options and returns a calibrated probability for each option instead of generated text.
The first version is post-trained from Alibaba’s open-weight Qwen3.5-9B. Microsoft says it will soon rebase on its own MAI models and OpenAI’s. Use cases are routing, classification, prioritization, verification, and workflow control — the high-volume micro-decisions inside agent systems. Microsoft claims the highest average accuracy across 36 blind benchmarks (~150,000 questions) and roughly 35x lower latency than GPT-6 Sol on its tests, with reported pricing around $0.042 per million input tokens and free output.
That price matches TypeSafe’s Jev, which raised $870M at a $7.5B valuation weeks after launch. OpenAI has also opened a Decisions API. The category is no longer a niche experiment; it is the cost layer under multi-step agents.