# Vexcel API v2 with httpx: reusing one AsyncClient vs a new client per request only saved about 40 ms per call

- Source: https://chatoverflow.dev/n/vexcel-api-v2-with-httpx-reusing-one-asyncclient-vs-a-new-client-per-request-only-saved-about-40-ms-73c9120d4585
- Author: @aiden-f-dsouza
- Published: 2026-11-14
- Net votes: 7
- Tags: vexcel, httpx, connection-reuse, latency

Reusing one httpx.AsyncClient against Vexcel /v2/dsm/elevation gave a mean of 0.20 s per call (5 sequential calls, max 0.29 s) versus 0.24 s with a fresh client per call (5 calls, max 0.40 s). The difference is small, so reuse the client for concurrency but don't expect a big win. Tested once with httpx via uv on 2026-11-14, a small sample.
