All Products
Search
Document Center

API Gateway:Gateway types

Last Updated:Aug 27, 2026

AI Gateway offers multiple instance types for different scenarios, from quick trials and development testing to production workloads and core production systems. This topic describes the positioning, capability differences, and selection recommendations for each instance type to help you choose the right one. For purchasable instance types, feature scope, billing items, pricing, service levels, and available regions, refer to the product console, the pricing page, and the Service-Level Agreement (SLA) page.

Instance type overview

The new Serverless edition is referred to collectively as Serverless and includes Serverless Standard and Serverless Enterprise. The original Serverless edition is referred to as Serverless (Legacy). The following table describes the positioning, scenarios, and key features of each instance type.

Instance typePositioningScenariosKey features
Serverless StandardEntry-level AI Gateway instanceQuick trials, proof of concept (PoC), development and testing, and low-traffic workloadsStarts at no cost. Pay-as-you-go. Uses a platform-assigned entry point.
Serverless EnterpriseProduction-grade Serverless AI Gateway instanceProduction workloads that require a dedicated entry point or custom domain namesProvides a dedicated entry point. Supports custom domain names and encryption-in-transit configuration.
Serverless (Legacy)Original Serverless editionContinued use by existing instances and smooth transitionExisting instances can continue to be managed and used. Capabilities remain at the original edition.
Dedicated InstanceHigh-specification dedicated instanceCore production, high concurrency, strong isolation, or workloads that require full governance capabilitiesProvides higher specifications, stronger isolation, and more complete gateway capabilities.

How to choose

The following table maps common scenarios to recommended instance types.

ScenarioRecommended instance typeDescription
Quick, low-cost experience of AI Gateway core capabilities such as intelligent routing, Agent API, and MCP ServerServerless StandardLow-cost trial. Suitable for low-frequency invocations and feature validation.
Pre-purchase PoC or technical validationServerless StandardQuick creation and low-cost validation. You can migrate to Serverless Enterprise later.
Development and testing environments that run long-termServerless StandardNo instance fee. Suitable for non-production environments.
Production workloads that require a higher service levelServerless EnterpriseProvides a dedicated entry point, a 99.95% service level, custom domain names, and certificate capabilities.
Custom domain names, certificates, TLS policies, or HTTP/2 configurationServerless EnterpriseThese settings are also available on Dedicated Instance.
SaaS services that provide a stable API entry point to external customersServerless EnterpriseA dedicated entry point and custom domain names fit external service requirements.
Existing Serverless (Legacy) production instances with no near-term change plansContinue using Serverless (Legacy) and evaluate migrationManagement operations remain unchanged. Plan migration within a business window.
Core production systems, high-compliance workloads, or workloads that require a 99.99% service levelDedicated InstanceProvides a 99.99% service level, stronger isolation, and full gateway capabilities.
High concurrency, strong isolation, or full advanced governance capabilitiesDedicated InstanceSuitable for heavy-load and complex governance scenarios.

Differences between the new Serverless edition and Serverless (Legacy)

The new Serverless edition includes Serverless Standard and Serverless Enterprise. It adopts a new architecture designed for low-cost onboarding, production-grade Serverless workloads, and continuous capability evolution. Serverless (Legacy) is the original Serverless edition, primarily intended for existing customers and transition scenarios. Existing Serverless (Legacy) instances can continue to be managed and used.

Compared with Serverless (Legacy), the new Serverless edition offers the following advantages:

  • Resource efficiency — The new Serverless edition uses a multi-tenant resource pooling architecture, which improves overall resource efficiency and better accommodates low-traffic and elastic workloads.

  • Provisioning and scaling — The new Serverless edition is better suited for workloads that require rapid creation, on-demand usage, and elastic growth. Serverless Standard starts at no cost, and Serverless Enterprise provides production-grade entry point capabilities.

  • Edition tiering — Serverless Standard and Serverless Enterprise target low-cost trials, development testing, and production workloads respectively. You can choose the right instance type for each business stage.

  • Capability evolution — The new Serverless edition hosts core AI Gateway capabilities such as LLM API, Agent API, MCP Server, intelligent routing, and observability analytics. New capabilities are delivered on the new Serverless edition first.

  • Plugin system — The new Serverless edition supports the platform-provided plugin system but not custom plugins. You can use plugins based on your instance type and feature availability to extend gateway capabilities such as authentication and authorization, security protection, traffic governance, and observability analytics.

Capability comparison

The following table compares the capabilities of all instance types.

Comparison itemServerless StandardServerless EnterpriseServerless (Legacy)Dedicated Instance
Recommended business stageTrials, evaluation, development testing, and low-traffic workloadsProduction and scaled workloadsProduction and scaled workloadsCore production, high concurrency, and high-compliance workloads
Starting costStarts at no costInstance fee appliesBase fee appliesBilled by dedicated specification
Billing methodPay-as-you-go based on actual usageInstance fee + pay-as-you-go based on actual usageBase fee + pay-as-you-go based on actual usageBilled by instance specification and actual usage
Access entry pointPlatform-assigned entry pointDedicated entry pointIndependent instance entry pointExclusive entry point
ArchitectureNew architecture, platform-assigned entry pointNew architecture, dedicated entry pointOriginal Serverless legacy architectureDedicated instance architecture
Plugin systemPlatform-provided plugin system. Custom plugins are not supported.Platform-provided plugin system. Custom plugins are not supported.No plugin systemFull plugin and advanced governance capabilities
Custom domain namesNot supportedSupportedSupportedSupported
Certificates, TLS, and HTTP/2Not supported. The platform default HTTPS entry point remains available.User-configurableUser-configurableSupported
Default QPS limit50 QPS500 QPSDepends on the original edition specificationDepends on the instance specification
Service level99.9%99.95%99.95%99.99%
AI core capabilitiesLLM API, Agent API, MCP Server, intelligent routing, AI security, Token usage dashboard, and other core capabilitiesAll Standard capabilities, plus more complete production-grade configuration capabilitiesCore capabilities of the original edition, including LLM API, Agent API, MCP Server, intelligent routing, AI security, and Token usage dashboardFull gateway capabilities
Advanced governanceSome advanced capabilities are unavailableMore enterprise-grade capabilities than Serverless StandardOriginal edition governance capabilitiesFull advanced governance capabilities

Serverless Standard

Serverless Standard is a low-cost way to experience AI Gateway quickly. Use Serverless Standard to validate core capabilities such as LLM API, Agent API, MCP Server, intelligent routing, AI security, and Token usage dashboard.

Scenarios

  • Individual developers who want to experience AI Gateway capabilities

  • Enterprises conducting pre-purchase PoC or technical validation

  • AI application startups onboarding AI Gateway during low-traffic phases

  • Independent software vendors (ISVs) or Software as a Service (SaaS) vendors that need development and testing environments

Key advantages

  • Starts at no cost — No instance fee. You pay based on actual usage.

  • Quick provisioning — Uses a platform-assigned entry point, which lowers the onboarding barrier.

  • Core AI capabilities — Supports core capabilities such as LLM API, Agent API, MCP Server, and intelligent routing.

Limits

  • QPS limit — The default instance-level QPS limit is 50.

  • Custom domain names — Associating custom domain names is not supported.

  • Certificates, TLS, and HTTP/2 — Custom certificates, TLS policies, and HTTP/2 configuration are not supported.

  • Entry point — No dedicated public network entry point is provided. The instance uses the platform-assigned entry point.

Serverless Enterprise

Serverless Enterprise is suitable for production workloads. Compared with Serverless Standard, Serverless Enterprise provides a dedicated entry point, a higher service level, and support for custom domain names and encryption-in-transit configuration.

Scenarios

  • AI API, Agent API, or MCP Server workloads that already carry production traffic

  • Enterprise workloads that require a stable, dedicated entry point

  • SaaS or ISV scenarios that need to provide a stable API entry point to downstream customers

  • Workloads that require custom domain names, certificates, TLS policies, or HTTP/2 configuration

Key advantages

  • Production-grade instance form — Suitable for production workloads.

  • Dedicated entry point — Meets production requirements for a stable entry point.

  • Higher service level — The service level is 99.95%.

  • Custom domain names — Supports associating custom domain names.

  • Encryption in transit — Supports user-configured certificates, TLS policies, and HTTP/2.

Limits

  • QPS limit — The default instance-level QPS limit is 500.

Serverless (Legacy)

Serverless (Legacy) is the original Serverless edition and uses the legacy architecture. It is primarily intended for existing customers and transition scenarios. Existing Serverless (Legacy) instances can continue to be managed and used. Available capabilities depend on the purchased instance specification. To purchase Serverless (Legacy), refer to the purchasable scope and product rules in the console.

If you are a new customer, choose Serverless Standard or Serverless Enterprise. The new Serverless edition uses a new architecture and is better suited for continuous AI Gateway capability upgrades in resource efficiency, provisioning flexibility, edition tiering, capability evolution, and plugin system support.

Scenarios

  • You already have Serverless (Legacy) instances and do not plan to change your workloads in the short term.

  • Your current workloads still depend on original edition capabilities, and migration requires more validation time.

  • You need to evaluate migration to the new Serverless edition or Dedicated Instance within a business window.

Limits

  • Plugin system — Not supported.

  • QPS limit — Depends on the original edition specification.

Dedicated Instance

Dedicated Instance is suitable for core workloads that require stronger resource isolation, higher service levels, higher performance specifications, or full governance capabilities.

Scenarios

  • Core production systems of large enterprises

  • Workloads that require a 99.99% service level

  • Workloads with long-term high concurrency or high burst traffic

  • Workloads that require more complete plugin, security, observability, and governance capabilities

  • Workloads with higher requirements for resource isolation, compliance, auditing, or network control

Key advantages

  • Stronger resource isolation — Suitable for core production and high-compliance scenarios.

  • Highest service level — The service level is 99.99%.

  • Full gateway capabilities — Suitable for complex plugin, security, observability, and governance requirements.

  • More stable performance expectations — Suitable for high-throughput, strong-isolation, and large-customer scenarios.

Limits

  • QPS limit — Depends on the instance specification.

Migration guidance

Evaluate the target edition based on your business stage, entry point requirements, service level, and governance needs.

From Serverless Standard to Serverless Enterprise

One-click upgrade from Serverless Standard to Serverless Enterprise is planned but not yet available. The upgrade will not require resource migration. After the upgrade, the instance gains Enterprise capabilities such as a dedicated entry point, custom domain names, encryption-in-transit configuration, higher QPS, and a higher service level. The exact timeline, upgrade conditions, and operation entry point are subject to the product console.

From Serverless (Legacy)

If you want to lower the entry barrier and quickly experience new AI capabilities, evaluate Serverless Standard. If you want a production-grade entry point, custom domain names, and a higher service level, evaluate Serverless Enterprise. If your instance carries core high-traffic workloads or depends on full advanced governance capabilities, evaluate Dedicated Instance.

From Serverless Enterprise to Dedicated Instance

Evaluate Dedicated Instance when your workload requires a higher service level, stronger resource isolation, high concurrency, full advanced governance capabilities, or stricter compliance.

FAQ

Is Serverless Standard suitable for production workloads?

Serverless Standard is suitable for low-traffic, cost-sensitive workloads that can use a platform-assigned entry point. If you need custom domain names, a dedicated entry point, a higher service level, or QPS beyond the default limit, choose Serverless Enterprise.

Does Serverless Standard support HTTPS?

The platform default HTTPS entry point is supported. Serverless Standard does not support custom certificates, TLS policies, or HTTP/2 configuration.

What is the difference between Serverless Enterprise and Dedicated Instance?

Serverless Enterprise is a production-grade Serverless edition that provides a dedicated entry point, custom domain names, encryption-in-transit configuration, and a 99.95% service level. Dedicated Instance is a high-specification dedicated instance. It suits core workloads that require stronger resource isolation, a 99.99% service level, higher performance specifications, or full plugin and advanced governance capabilities.

Should new customers choose Serverless (Legacy)?

Serverless (Legacy) is primarily intended for existing customers and specific transition scenarios. If you are a new customer, choose Serverless Standard or Serverless Enterprise.