DEX Latency Map
Public-network latency measurements to the order APIs of decentralized exchanges (DEXs). This tab covers order-book DEXs that run their own matching engine.
These numbers measure what a public API client experiences reaching the endpoint each DEX takes orders on (REST time to first byte — see the "Measured" column below). They're useful for deciding where to host bots and which region gives the lowest network access latency. They are not a prediction of end-to-end order execution time. Hyperliquid has its own tab.
What is measured
Every probe sends a request to each DEX's order endpoint without credentials. The DEX rejects it with its own authentication or validation error, and the time until that answer arrives is recorded. A sample only counts when the response is the DEX's own rejection, so an answer from a CDN edge or a regional block never shows up as latency.
This is the same path a signed order takes up to the DEX's API gateway. The last hop inside the DEX, from the gateway to the matching engine, is internal and not part of the number.
Off-chain matching, on-chain settlement
None of these DEXs matches orders on a blockchain. Each runs its own order engine and settles results on-chain afterwards. Orders reach that engine through its API rather than through a smart contract, so where the order API answers from is what decides latency.
Lighter on Robinhood Chain shows how far apart those two things can be. Its order API answers from Tokyo, while Robinhood Chain itself is sequenced in the US and only receives a batch of results about once a minute.
Where these DEXs run
Every DEX on this tab answers fastest from our Tokyo probes, the same city Hyperliquid runs from. Latency rises steadily with distance: a few milliseconds inside Tokyo, tens of milliseconds from Seoul and Hong Kong, and over two hundred from Europe.
Why some DEXs are missing from some probes
Extended, Lighter and Lighter on Robinhood Chain refuse requests from the US and UK, so probes in Chicago, Ashburn, San Jose, Ohio and London show no data for them. Extended also blocks Hong Kong, Lighter on Robinhood Chain also blocks Singapore, and all three block Johannesburg. A probe in a restricted region shows no data for that DEX rather than the time it takes to be turned away.
Why some endpoints don't show TCP/TLS breakdown
Most of these DEXs sit behind a CDN or an anycast front end, where the TCP and TLS handshakes end at the nearest edge rather than at the DEX. We hide the TCP/TLS columns for those endpoints. The only honest number for them is the p50 round-trip, which travels all the way from the probe through the edge to the origin and back.
Extended is the exception: it answers directly from AWS Tokyo, so its TCP and TLS times reflect real probe→origin distance and are shown.
Probable Origin: Appears to run in Tokyo
Orders are sent to Aster's own engine over REST, with settlement on-chain. Spot orders use a separate API that answers from the same Tokyo location. Fronted by CloudFront; every order request is passed through to the origin.
Endpoints
| Role | Endpoint | Host | Measured | Routing |
|---|---|---|---|---|
| Primary | Order API | fapi.asterdex.com/fapi/v1/order | REST POST TTFB | CloudFront |
Research Project Disclaimer
DEX Latency Map is a research project provided for informational and educational purposes only. The numbers measure network distance to each DEX's order gateway, not end-to-end order-submission or execution time. Actual order round-trip times depend on additional factors including matching-engine processing, rate limits and account tier. These numbers should be viewed as estimates and directional baselines.