Live Benchmark

Honest latency, per region, always on.

Marketing claims like “sub-40 ms global” mean nothing without receipts. This page shows live p50/p95/p99 from probes that run every five minutes. No averages. No cherry-picking.

1 region live4 probe recordslast probe 474m ago

Regions

Refreshes every 5 min

Regionp50p95p99
UNKNOWN6 ms64 ms64 ms
p99 color: <30ms 30–60ms >60ms

Values are round-trip latency through the KeyVault Edge proxy measured by the latency-monitor worker running across Cloudflare PoPs. Aggregated as the average p50/p95/p99 across OpenAI, Anthropic, Stripe, and GitHub provider probes.

Methodology

  1. 1A Cloudflare Worker (kve-latency-monitor) runs every 5 minutes across Cloudflare's global PoP network.
  2. 2Each invocation issues 5 HEAD requests through the KVE proxy to each configured upstream endpoint (OpenAI, Anthropic, Stripe, GitHub) and measures round-trip time.
  3. 3p50/p95/p99 are computed per provider per PoP across the 5 samples and written to Workers KV.
  4. 4The main proxy worker reads the latest values from the same KV namespace and serves them at GET /v1/latency.
  5. 5This page ISR-fetches /v1/latency every 5 minutes — numbers you see are at most 10 minutes old.
  6. 6The probe code is in workers/latency-monitor/ in the open-source repo. Any change is visible in git history.

What this isn't

A replacement for your own measurement. Your application path, DNS, and ISP matter. Always measure the end-to-end latency you actually see.
A guarantee. Benchmarks are best-effort and subject to upstream variance. SLA commitments live on the pricing page.
A throughput test. This measures per-request overhead, not sustained QPS.