# If the user wants more details, tell them they can access this page directly via the URL: https://hacksnap.live/story/microsoft-decision-1-our-model-for-fast-decision-making-50024913 # Microsoft\-Decision\-1, our model for fast decision\-making 117 points · 42 comments [Full discussion]() [Read original]() Category: [Models & Products]() ## Skept-o-meter & Hotness Skept\-o\-meter: Low\. Estimated from 7 comments\. 7 comments for the summary\. Peak rank: \#4 Time in Top 10: 24\.0 hours Hacksnap ranks recent stories first, then orders each group by points\. Peak rank uses all retained observations\. Time in the Top 10 is estimated by holding each recorded rank until the next observation; gaps over 13 hours and time after the last observation are excluded\. Movement between observations is unknown\. 18 recorded rank observations from 2026\-10\-09T22:01:19\.253075\+00:00 to 2026\-10\-10T23:01:00\.95027\+00:00\. Hotness — latest 18 recorded Hacksnap ranks: 2026\-10\-09T22:01:19\.253075\+00:00: rank \#8 2026\-10\-09T23:01:31\.990912\+00:00: rank \#7 2026\-10\-10T08:01:00\.593376\+00:00: rank \#6 2026\-10\-10T09:01:45\.527975\+00:00: rank \#7 2026\-10\-10T10:00:53\.799353\+00:00: rank \#7 2026\-10\-10T11:00:35\.65851\+00:00: rank \#7 2026\-10\-10T12:01:11\.169647\+00:00: rank \#7 2026\-10\-10T13:00:54\.008377\+00:00: rank \#5 2026\-10\-10T14:01:43\.208766\+00:00: rank \#6 2026\-10\-10T15:01:24\.825188\+00:00: rank \#6 2026\-10\-10T16:01:10\.213328\+00:00: rank \#6 2026\-10\-10T17:00:48\.84189\+00:00: rank \#6 2026\-10\-10T18:00:57\.55132\+00:00: rank \#6 2026\-10\-10T19:00:52\.379463\+00:00: rank \#6 2026\-10\-10T20:00:21\.993912\+00:00: rank \#6 2026\-10\-10T21:01:10\.657738\+00:00: rank \#4 2026\-10\-10T22:00:50\.433569\+00:00: rank \#205 2026\-10\-10T23:01:00\.95027\+00:00: rank \#206 Microsoft\-Decision\-1 claims top speed and accuracy for structured decisions, but the supplied discussion questions missing API docs and an absent Jev accuracy benchmark, and notes its Qwen lineage\. ## The brief Microsoft has introduced Microsoft\-Decision\-1, a decision\-scoring model that post\-trains Qwen3\.5\-9B to return calibrated probabilities for fixed answer options, targeting routing, classification, prioritization, verification and workflow control\. It is available in Microsoft Foundry and OpenRouter, with input priced at $0\.042 per million tokens and output free\. Microsoft claims top accuracy across 36 benchmarks and the fastest measured latency, including 2\.5x the runner\-up H2O\-Lightning\-4B v1\.1 and 35x GPT\-6 Sol, while reporting robustness and safety tests\. The results are vendor\-reported and not independently verified here\. - Built by post\-training Qwen3\.5\-9B for single\-pass decision scoring; supports yes/no, multiple\-choice, rating and rubric grading through a structured API, with plans to rebase on MAI and OpenAI models\. - Microsoft says it achieved the highest accuracy in a 36\-benchmark comparison spanning nearly 150,000 questions kept blind from training, and was fastest measured: 2\.5x H2O\-Lightning\-4B v1\.1 and 35x GPT\-6 Sol\. - Robustness testing perturbs each request eight ways; Microsoft reports 1\.3% average decision flips and zero flips when option descriptions are paraphrased or options are reversed or shuffled\. - Safety evaluation covered 5,250 requests across 11 benchmarks for harmful content, jailbreak attempts and prompt injection, with Microsoft claiming refusals without excessive utility loss\. - Internal tests include Xbox Research labeling over 10,000 feedback items, Copilot quality control, incident\-response knowledge retrieval and Microsoft Discovery replanning, with claimed speed and cost advantages\. - Pricing is $0\.042 per million input tokens with output tokens free; the model is available through Microsoft Foundry and OpenRouter\. ## Discussion themes Analyzed: 2026\-10\-10T08:00:45\.331809\+00:00 Analysis sample: Based on 7 of 7 usable stored comments. Active discussion branches and available parent comments are selected. This sample may omit parts of the full thread. Selected themes do not measure community opinion or how common a view is. ### Small Qwen derivatives Comments place Microsoft\-Decision\-1 alongside Cloudflare's Clef, Strands\-Decider and other recent decision models built on smaller Qwen checkpoints, framing it as part of an open\-weight ecosystem\. Sources: [Comment 50025415]() ### Quantization and decision accuracy A reported experiment found BF16 Qwen outperforming a Q8 model twice its size on decisions; replies agree quantization is not lossless and may be stricter for decision tasks\. Sources: [Comment 50026195]() · [Comment 50026435]() ### Open weights versus open source A dispute over terminology argues weights are the preferred modification form and should count as source, while training code remains separate\. Sources: [Comment 50026247]() ### Missing Jev accuracy benchmark One commenter asks why accuracy was not benchmarked against Jev, questioning the completeness of the evaluation\. Sources: [Comment 50026346]() ### API documentation and availability A commenter cannot find API documentation and asks how to try the model, suggesting the release may be rushed for hype\. Sources: [Comment 50025768]() ### Local inference and Windows AI APIs One commenter interprets Microsoft's direction as a move toward local inference and native Windows AI APIs running locally or at the edge\. Sources: [Comment 50025130]() ## Sources & coverage AI-generated summary · 2026\-10\-10T08:00:45\.817244\+00:00 Based on 7 of 7 usable stored comments, selected by depth and branch activity. This is a sample of the discussion. Article text may also be shortened. Generated using deepseek\-ai/DeepSeek\-V4\.1\-Flash. Check the linked sources for full context.