LLM gateway comparison
Compare fees, model access, BYOK, routing, observability, and documented country or region limits across four LLM gateways.
Compare LLM gateways
An LLM gateway puts model access, routing, credentials, usage controls, and logs behind one API. Provider inference charges and gateway fees remain separate costs.
| Gateway | Published models | Gateway pricing | BYOK | Routing | Observability | Availability and data controls |
|---|---|---|---|---|---|---|
 OpenRouterA strong fit when one API needs broad model and provider coverage. | 421In our latest official API snapshot | 5.5% when purchasing creditsProvider rates deducted from prepaid USD credits | Yes. Plan-based allowance, then a 5% fee | Provider order, fallbacks, price and privacy controls | Request metadata by default; prompt logging is opt-in | OpenRouter does not publish a definitive list of supported user countries.Check country and region limits |
 Vercel AI GatewayA practical fit for teams already using Vercel or the AI SDK. | 243In our latest official API snapshot | 0% token markupProvider list price | Yes. Dashboard or request-scoped, with no gateway fee | Automatic provider choice, ordering and fallbacks | Usage and spend through the Vercel dashboard | Vercel does not publish a country-by-country availability list for AI Gateway.Check country and region limits |
 Cloudflare AI GatewayA practical fit for Cloudflare users who need caching and edge controls. | 28In our latest official API snapshot | 5% unified billing feeProvider rates pass through | Yes. Provider keys can be stored in Secrets Store | Dynamic routing, fallbacks, rate limits and caching | Analytics, configurable logging and Logpush export | Cloudflare documents AI Gateway as available on all Cloudflare plans, but does not publish a country-by-country user availability list for the service.Check country and region limits |
 PortkeyA strong fit for teams that need self-hosting or deeper operational controls. | Not collectedNo authenticated catalog snapshot | $49/mo ProductionBYOK and provider billing | Yes. Provider integrations and privately hosted models | Conditional routing, fallbacks, retries and load balancing | Logs, traces, costs and configurable retention | Portkey does not publish a country-by-country availability list for its managed AI Gateway.Check country and region limits |
Catalog counts come from official APIs checked on August 22, 2026. A larger catalog is not automatically a better choice. Confirm model, provider, and data-policy availability for your workload.
The platform pages distinguish account availability, model-provider restrictions, data residency, and edge coverage. They do not infer country support from currencies or network locations.
How to choose the best LLM gateway for your setup
OpenRouter
A strong fit when one API needs broad model and provider coverage.
Vercel AI Gateway
A practical fit for teams already using Vercel or the AI SDK.
Cloudflare AI Gateway
A practical fit for Cloudflare users who need caching and edge controls.
Portkey
A strong fit for teams that need self-hosting or deeper operational controls.
Official sources
Pricing and product capabilities change. Check the linked provider documentation before making a purchasing or compliance decision.