Microsoft-Decision-1: fast decision-scoring model for structured choices
Microsoft's purpose-built decision model that returns calibrated probabilities for fixed option sets instead of generating text.
Why it matters
First hyperscaler decision model: 4.5x faster than its closest rival on Microsoft's 36-benchmark eval, at $0.042/M input tokens with free outputs.
What you could build with it
A support-ops startup could use it as a cheap deterministic triage gate: every incoming ticket gets classified and priority-scored in one low-latency pass, with only low-confidence cases escalated to a full LLM, cutting routing inference spend by an order of magnitude.
Does it hold up?
Too early to judge from independent use: Microsoft reports internal wins (Xbox Research categorized 10k+ feedback items at 14x the speed and 200x lower cost vs GPT-6 Sol), but third-party benchmarks and production deployments are not yet public.
Built with Microsoft-Decision-1
- Microsoft launches Decision-1 model in FoundryTestingCatalog · Independent writeup of the launch: pricing, yes/no and multiple-choice and rating support, and the 36-benchmark accuracy and latency claims.
- Microsoft-Decision-1 Is Here With 35x Faster Performance Than GPT-6 SolWindowsReport · Coverage of the launch including the Qwen3.5-9B post-training base and internal use by Xbox Research and the Copilot team.
- Microsoft-Decision-1 on OpenRouterOpenRouter · Live API listing confirming the Oct 9 release date, $0.042/M input pricing, zero output cost, and the Decisions-API request shape.
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74Mistral Large 4 (Le Chonk): 1.05T-param open-weight MoEMistral's new flagship: a 1.05T-parameter mixture-of-experts multimodal model with 49B active params and 1M-token…models · JEV 0.72ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71JetBrains Mellum 2.1: 12B MoE coding model trained for agentsJetBrains released Mellum 2.1, a 12B MoE model (2.5B active parameters, 128K context) under Apache 2.0, trained mainly…models · JEV 0.69DeepSeek V4-Flash official API: public beta with upgraded agent capabilities, Responses API and Codex supportDeepSeek's official V4-Flash is now a public beta API with massively upgraded agent capabilities - benchmark scores…models · JEV 0.69
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.