| title | Latency and Server Locations |
|---|---|
| description | API gateway regions and colocation guidance for lowest latency |
| llmsDescription | Jupiter's API gateway server locations and colocation guidance for integrators optimising latency. The API gateway runs on AWS in six regions: Singapore (ap-southeast-1), Frankfurt (eu-central-1), Virginia (us-east-1), São Paulo (sa-east-1), Tokyo (ap-northeast-1), and Oregon (us-west-2). Covers how to choose the closest region for your servers, general colocation recommendations, and how latency varies by proximity. Applies to all Jupiter API products accessed via api.jup.ag and lite-api.jup.ag. |
Jupiter's API gateway is distributed across six AWS regions worldwide. Latency to the API depends primarily on the physical distance between your servers and the nearest gateway region. Colocating your servers in or near one of these regions gives you the lowest possible round-trip time.
| Region | AWS Region ID | Location |
|---|---|---|
| Singapore | ap-southeast-1 |
Southeast Asia |
| Frankfurt | eu-central-1 |
Europe |
| Virginia | us-east-1 |
US East |
| São Paulo | sa-east-1 |
South America |
| Tokyo | ap-northeast-1 |
East Asia |
| Oregon | us-west-2 |
US West |
All regions serve the same API surface with the same data freshness. There is no difference in functionality between regions.
For the lowest API latency, deploy your servers in the same AWS region (or the same availability zone) as one of the gateway regions listed above. If your infrastructure runs on a different cloud provider, choose a data centre in the same city or metro area.
If your users are globally distributed:
- Asia-Pacific: Singapore (
ap-southeast-1) or Tokyo (ap-northeast-1) - Europe: Frankfurt (
eu-central-1) - Americas: Virginia (
us-east-1), Oregon (us-west-2), or São Paulo (sa-east-1)
If you want to verify quote-path latency from your own region before colocating, OpenChainBench publishes an open, multi-region latency benchmark for Solana DEX aggregators. It probes the quote API on Jupiter, Raydium, OpenOcean and Mobula every 60 seconds from Paris, Virginia and Singapore, and tracks p50 / p95 wall-clock latency plus response success rate. Methodology and raw data are open-source (source).