Understanding deployment types in Microsoft Foundry Models - Microsoft Foundry | Microsoft Learn (https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/deployment-types ; rewritten summary of https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/deployment-types) citeturn142view1 [wordlim: 200] Crawled: today; Content type: text/html; Source: open({"ref_id":"https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/deployment-types","lineno":null}); Redirected to URL: https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/deployment-types; Total lines: 271 L0: 【0†Skip to main content】 【1†Skip to Ask Learn chat experience】 L1: L2: This browser is no longer supported. L3: L4: Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support. L5: L6: 【2†Download Microsoft Edge†go.microsoft.com】 【3†More info about Internet Explorer and Microsoft Edge】 L7: L8: Table of contents Exit editor mode L9: L10: Ask Learn Ask Learn L11: L12: Reading mode Table of contents 【1†Read in English】 Add Add to Plans 【4†Edit†github.com】 L13: L14: * * * L15: L16: Copy Markdown Print L17: L18: * * * L19: L20: Note L21: Access to this page requires authorization. You can try 【1†signing in】 or changing directories. L22: L23: Access to this page requires authorization. You can try changing directories. L24: # Deployment types for Microsoft Foundry Models L25: L26: Feedback L27: L28: Summarize this article for me L29: ## In this article L30: L31: When you deploy a model in Microsoft Foundry, you choose a deployment type that determines: L32: L33: * Where your data is processed (global, data zone, or Azure geography) L34: * How you pay (pay-per-token or reserved capacity) L35: * Performance characteristics (latency variance, throughput limits) L36: These deployment types apply to the Serverless API deployment option. Open-source and custom models that use managed compute don't use these types. For how the options differ, see 【5†Deployment overview for Microsoft Foundry Models】. L37: The service offers three main categories: standard (pay-per-token), provisioned (reserved capacity), and batch (discounted asynchronous processing). A Developer type is also available for fine-tuned model evaluation. Within the standard and provisioned categories, you can choose global, data zone, or geography-based processing based on your compliance requirements. L38: L39: Tip L40: You don't always need to create a deployment. With 【6†instant access (preview)】, you call supported models by name and start running inference immediately — no deployment required. L41: L42: Important L43: L44: Data residency for all deployment types: Data stored at rest remains in the designated Azure geography. However, inferencing data is processed as follows: L45: * Global types: May be processed in any Azure region L46: * Data Zone types: The service processes data only within the Microsoft-specified data zone (US, EU, or Asia Pacific (APAC)). L47: * Standard and Regional Provisioned types: Prompts and responses are processed within the customer-specified Azure geography and might be processed between regions within that geography for operational purposes. L48: L49: 【7†Learn more about data residency†azure.microsoft.com】. L50: ## Start with Global Standard L51: L52: For most workloads, start with Global Standard. It launches first when a new model releases, has the lowest price, and offers the broadest region coverage. Move to another deployment type only when you have a specific reason, such as data residency, reserved throughput, or asynchronous batch processing. L53: New deployment types become available in a set order: Global, then Data Zone, then geography-based. Geography-based deployment types arrive last, have no guaranteed availability date, and depend on capacity that frees up as older models retire. For the authoritative launch order, see 【8†Model launch and availability】. L54: ## Deployment type comparison L55: L56: Instant models let you run inference without creating a deployment, so they aren't deployment types. To try a model instantly, see 【6†Instant access to models】. L57: Deployment type | SKU code | Data processing | Billing | Best for L58: --- | --- | --- | --- | --- L59: 【9†Global Standard】 | `GlobalStandard` | Any Azure region | Pay-per-token | General workloads, highest quota L60: 【10†Global Provisioned】 | `GlobalProvisionedManaged` | Any Azure region | Reserved PTU | Predictable high-throughput L61: 【11†Global Batch】 | `GlobalBatch` | Any Azure region | 50% discount, 24-hr | Large async jobs L62: 【12†Data Zone Standard】 | `DataZoneStandard` | Within data zone | Pay-per-token | EU/US/APAC data zone compliance L63: 【13†Data Zone Provisioned】 | `DataZoneProvisionedManaged` | Within data zone | Reserved PTU | Data zone + predictable throughput L64: 【14†Data Zone Batch】 | `DataZoneBatch` | Within data zone | 50% discount | Large async jobs with data zone L65: 【15†Standard】 | `Standard` | Within Azure geography | Pay-per-token | Geography compliance, low volume L66: 【16†Regional Provisioned】 | `ProvisionedManaged` | Within Azure geography | Reserved PTU | Geography compliance + throughput L67: 【17†Developer】 | `DeveloperTier` | Any Azure region | Pay-per-token | Fine-tuned model evaluation only (24-hour lifetime, no SLA or data-residency guarantee) L68: Note L69: L70: Not all models support all deployment types. Check 【18†Foundry Models sold by Azure】 for model availability by deployment type and region. L71: L72: SLA guarantees vary by deployment type. Provisioned types provide guaranteed throughput and lower latency variance. Standard types offer best-effort service. Developer deployments don't include an SLA. For details, see the 【19†Azure SLA for Azure OpenAI Service†www.microsoft.com】. L73: L74: Tip L75: For detailed pricing, see 【20†Azure OpenAI Service pricing†azure.microsoft.com】. L76: ## Choose the right deployment type L77: L78: Use the following table for a quick recommendation, then refine it with the criteria that follow. L79: Requirement | Recommended tier L80: --- | --- L81: Default: newest models, lowest price, broadest regions | Global Standard L82: Reserved, predictable throughput | Global Provisioned L83: Keep processing within a data zone | Data Zone Standard or Data Zone Provisioned L84: Data residency plus reserved throughput | Data Zone Provisioned L85: Keep processing within an Azure geography | Standard or Regional Provisioned (where supported) L86: Large asynchronous jobs at lower cost | Global Batch or Data Zone Batch (where supported) L87: Evaluate a fine-tuned model (temporary, no SLA) | Developer L88: ### By data residency requirement L89: L90: * No restrictions: Use Global Standard or Global Provisioned L91: * EU, US, or APAC data zone: Use Data Zone Standard or Data Zone Provisioned in a region within that data zone L92: * Azure geography: Use Standard or Regional Provisioned L93: L94: For the exact regions in each data zone, see 【21†Data Zone deployments】. L95: ### By workload pattern L96: L97: * Quick start, prototyping, or trying a new model: Use 【6†instant access (preview)】 (no deployment needed) L98: * Variable, bursty traffic: Use Standard or Global Standard (pay-per-token) L99: * Consistent high volume: Use Provisioned types (reserved capacity)