Science & Tech · AI · 7 hrs ago
Microsoft releases a fast model for scoring AI decisions
Microsoft has released Microsoft-Decision-1, a model designed to score choices rather than generate open-ended text.
It takes a situation, a question and set answer options, then gives each option a probability.
Developers can use those scores to classify information, prioritize tasks or help an AI agent choose its next step, and can set it to defer or request a review when confidence is low.
Microsoft says it tested the model on 36 benchmarks and that it was faster than the other models it compared it with.
In its own tests, Microsoft also reported faster processing on tasks such as sorting Xbox feedback and assessing AI responses.
The model is not intended for conversation, translation or summarization, and Microsoft advises developers to test it on examples that match their intended use.
It is available through Microsoft Foundry and OpenRouter for US$0.042 per million input tokens, with no charge for output tokens.
Microsoft has released Microsoft-Decision-1, a model that scores fixed answer options with calibrated probabilities.
The company says it is designed to help applications and AI agents route, classify, prioritize, verify and control workflows.
Microsoft reported that the model was 2.5 times faster than the next-fastest model in its benchmark and about 35 times faster than GPT-6 Sol in median latency.
The model is generally available through Microsoft Foundry and OpenRouter, at US$0.042 per million input tokens.
Microsoft says the model is not designed for open-ended text generation, conversation, translation or summarization.
- Who
- Microsoft released Microsoft-Decision-1.
- What
- A model that scores a fixed set of answer options and returns a probability for each.
- When
- It is generally available; a release date is not stated.
- Where
- Through Microsoft Foundry and OpenRouter.
- Why
- It is designed to support routing, classification, prioritization, verification and workflow control.
This story does not have two clearly opposing sides.
This model is designed for routing, classification, prioritization, verification, and workflow control, making it easier to incorporate decision intelligence into existing applications, agents, and workflows in a secure, trusted environment,
This story does not have a timeline yet.
- Benchmark scope
- 36 benchmarks covering nearly 150,000 questions
- Speed comparison
- 2.5 times faster than H2O-Lightning-4B v1.1 and about 35 times faster than GPT-6 Sol in median latency
- Availability
- Microsoft Foundry and OpenRouter
- Input price
- US$0.042 per million tokens
- Output price
- No charge
Sources
Related news
Microsoft president calls for global AI safety rules
Nadella calls for human-controlled emergency brakes on advanced AI
Meta releases its first frontier AI models, Muse Wonder and Wonder Full
CNN asks readers to share questions about AI safety risks
Meta launches paid Muse Wonder AI model for its personal agent app
AI agent accessed Australian government systems during research task
Anthropic AI submits false murder tip to Philadelphia police
Meta Unveils Muse Wonder AI Models to Compete With Leading Systems
Meta Unveils Two Advanced AI Models
Zuckerberg says Meta has reached AI frontier and rejects slowdown calls
Meta launches Muse Wonder and Wonder Full AI models
