How much does Microsoft-Decision-1 cost?

Microsoft-Decision-1 costs $0.042 per million input tokens, and output tokens are free, according to Microsoft's launch post. Because the model returns probabilities for a fixed set of options rather than generated text, almost all of the cost of a call comes from the input state and the question being scored.

How fast is Microsoft-Decision-1 compared with other models?

Microsoft says Microsoft-Decision-1 had the lowest latency it measured: its P50 latency is about 35x faster than GPT-6 Sol and 2.5x faster than H2O-Lightning-4B v1.1, the runner-up. In Copilot team tests it was competitive with GPT5.6 Luna while running 100x faster. These are Microsoft's own benchmark results.

Is Microsoft-Decision-1 open source?

Microsoft-Decision-1 is built on Alibaba's open-weight Qwen3.5-9B, but Microsoft's launch post does not say that its own weights are released, and it links no Hugging Face or GitHub page for the model. Developers reach it as a hosted model in Microsoft Foundry, where the catalog lists it as Version 1, direct from Azure.

What should Microsoft-Decision-1 not be used for?

The Microsoft Foundry catalog says Microsoft-Decision-1 is not meant for open-ended generation, conversation, translation or summarization. It also says the model should not be the sole decision-maker in consequential decisions about people. Its listed uses are AI-output evaluation, classification and routing, agent guardrails, content-safety screening and document relevance.

How was Microsoft-Decision-1 tested for safety?

Microsoft tested Microsoft-Decision-1 on 5,250 requests across 11 benchmarks covering harmful content, jailbreaks and prompt injection. For accuracy, it reports the highest score in a 36-benchmark comparison of nearly 150,000 questions held out from training, against models such as Quyet-1.0-Large, deck-31B, GPT-6 Luna Decisions and Strands Decider 2B.

Microsoft · 2026-10-09 · major

Microsoft-Decision-1 — a fast decision model at $0.042 per million tokens

Microsoft-Decision-1 is a decision-scoring model post-trained from Qwen3.5-9B. It returns a calibrated probability for each answer option, runs about 35x faster than GPT-6 Sol at P50, and costs $0.042 per million input tokens.

Microsoft's first decision model scores fixed answer options in one pass, for routing, guardrails and grading agent actions.

Key specs

Speed vs gpt 6 sol (p50)	~35x faster
Benchmarks compared	36

Quick facts

Maker	Microsoft
Base model	Qwen3.5-9B, post-trained by Microsoft
Price (input)	$0.042 / 1M tokens
Price (output)	Free
Context window	32,768 tokens
Availability	Microsoft Foundry
Output	Calibrated probability per answer option (JSON)

Pricing

Input	$0.042 / 1M tokens
Output · output tokens are free	$0 / 1M tokens

source ↗

What is it?

Microsoft-Decision-1 is a decision-scoring model that Microsoft announced on October 9, 2026. Instead of writing text, it takes a fixed set of answer options — yes/no, multiple choice or a rating — and returns a calibrated probability for each one. It can also grade AI responses and agent actions against a rubric.

How does it work?

Microsoft post-trained Alibaba's open-weight Qwen3.5-9B for fast, single-pass decision scoring, served behind a structured API call. The Foundry catalog lists a 32,768-token context, text input and JSON output. In Microsoft's tests the model changed its answer on 1.3% of perturbations on average, and never flipped when option descriptions were paraphrased or options were reversed or shuffled.

Why does it matter?

Agent pipelines make many small decisions — which tool to call, which team gets a ticket, whether an output is safe — and paying a large LLM for each one is slow and costly. Microsoft reports that on XBOX Research feedback triage the model matched GPT-6 Sol's quality at more than 14x the speed and 200x lower cost. It also says it will soon rebase the model on MAI and OpenAI models.

Who is it for?

developers building routing, classification and guardrail steps for agents

Frequently asked questions

How much does Microsoft-Decision-1 cost?: Microsoft-Decision-1 costs $0.042 per million input tokens, and output tokens are free, according to Microsoft's launch post. Because the model returns probabilities for a fixed set of options rather than generated text, almost all of the cost of a call comes from the input state and the question being scored.
How fast is Microsoft-Decision-1 compared with other models?: Microsoft says Microsoft-Decision-1 had the lowest latency it measured: its P50 latency is about 35x faster than GPT-6 Sol and 2.5x faster than H2O-Lightning-4B v1.1, the runner-up. In Copilot team tests it was competitive with GPT5.6 Luna while running 100x faster. These are Microsoft's own benchmark results.
Is Microsoft-Decision-1 open source?: Microsoft-Decision-1 is built on Alibaba's open-weight Qwen3.5-9B, but Microsoft's launch post does not say that its own weights are released, and it links no Hugging Face or GitHub page for the model. Developers reach it as a hosted model in Microsoft Foundry, where the catalog lists it as Version 1, direct from Azure.
What should Microsoft-Decision-1 not be used for?: The Microsoft Foundry catalog says Microsoft-Decision-1 is not meant for open-ended generation, conversation, translation or summarization. It also says the model should not be the sole decision-maker in consequential decisions about people. Its listed uses are AI-output evaluation, classification and routing, agent guardrails, content-safety screening and document relevance.
How was Microsoft-Decision-1 tested for safety?: Microsoft tested Microsoft-Decision-1 on 5,250 requests across 11 benchmarks covering harmful content, jailbreaks and prompt injection. For accuracy, it reports the highest score in a 36-benchmark comparison of nearly 150,000 questions held out from training, against models such as Quyet-1.0-Large, deck-31B, GPT-6 Luna Decisions and Strands Decider 2B.

Try it

https://ai.azure.com/catalog/models/Microsoft-Decision-1