Skip to content
llm-spend

Microsoft Decision Models

Microsoft

Microsoft-Decision-1 scores bounded choices at $0.042 per million input tokens.

Microsoft-Decision-1 is a Foundry model for classification, routing, ranking, and evaluation. It returns calibrated scores for fixed answer options rather than generated prose, and supports a 32K-token context.

The Foundry blog lists $0.042/M input for US and EU Data Zone deployments, with output listed as N/A; Microsoft's developer article says output tokens are free. The Azure Retail Prices API lists a $0.042/M Global meter effective October 1. Microsoft Learn documents both Global Standard and Data Zone Standard deployment types. This decision-scoring model is listed separately from generative workload comparisons.

Pricing

per 1M tokens · USD / CHF
ModelTier / HostContextInputCachedOutputConfidence
Microsoft-Decision-1
Decision-scoring model for predefined choices; it returns structured probability scores instead of generated text. Microsoft lists output tokens as free.
Foundry ·Global
Microsoft Foundry
32.8K$0.042CHF 0.034—$0.00CHF 0.00official
Azure Retail Prices API, serviceName 'Foundry Models', captured 2026-10-11: the 37-region Global meter 'Microsoft Decision 1 1M Tokens' is $0.042 per 1M tokens, effective 2026-10-01; its SKU metadata is 'Pi 1 glbl'. Microsoft's developer article says output tokens are free. Microsoft Learn documents GlobalStandard and DataZoneStandard deployments. The Microsoft Foundry catalog lists a 32,768-token context and JSON output.
Microsoft-Decision-1
Microsoft's launch table lists the same $0.042/M input rate for US and EU Data Zone deployments; the accompanying developer article says output tokens are free.
Foundry ·Data Zone
Microsoft Foundry
32.8K$0.042CHF 0.034—$0.00CHF 0.00official
Microsoft Foundry Blog, published 2026-10-09 and checked 2026-10-11, lists Microsoft-Decision-1 at $0.042 per 1M input tokens in both US Data Zone and EU Data Zone, with output listed as N/A. Microsoft's developer article states output tokens are free. Microsoft Learn documents DataZoneStandard as a supported deployment type. The Foundry catalog lists a 32,768-token context and JSON output. The Azure Retail Prices API separately lists only a Global meter for this model as of 2026-10-11, so the Data Zone amount is sourced to Microsoft's launch table rather than inferred from that meter.
Confidenceofficial published pagederived reconciled from billing †estimate pattern-inferred ‡
Other providers
KimiDeepSeekGLMOpenAI / Azure OpenAIOpenAI Decisions APICohereClaudeGeminiGrokQwenMistralMiniMaxPerplexityMicrosoft AIEmbeddingsCompare all →