Roar AI Cloud

AI gateway

Same model. Shorter trip.

Every major model, behind one OpenAI-compatible endpoint. Switch regions to watch the network cost change — the model's thinking time stays the same wherever you call it from, so the round trip is the part proximity actually buys.

Pick a region and send a prompt. The response streams token by token — the counters on the right update as it arrives.

This request

gateway round trip
–– ms
time to first token
–– ms
total time
–– ms
≈ tokens / second
––
characters
––

Routed through api.lk.roar-ai.com using the standard OpenAI SDK — only the base URL differs.

Gateway round trip is the network cost of reaching this region — that is what proximity buys you. Time to first token adds the model's own thinking time, which is the same wherever you connect from.