AI API latency, week of 24 August 2026: what actually moved
· covers 24 August 2026
Week of 24 August 2026: 4 different providers were fastest across the 4 regions we measure from, and 9 provider-region pairs moved by more than a quarter against the week before.
Who led each region last week?
| Region | Fastest | p50 TTFB | Uptime |
|---|---|---|---|
| Asia (Tokyo) | fireworks | 16 ms | 100% |
| Europe (Germany) | nscale | 98 ms | 100% |
| South America (São Paulo) | openrouter | 58 ms | 100% |
| US (Central) | 46 ms | 100% |
What changed by more than 25%?
| Provider | Region | Week before | This week | Change |
|---|---|---|---|---|
| reka | Europe (Germany) | 102 ms | 201 ms | +97% |
| glm | Asia (Tokyo) | 290 ms | 88 ms | -70% |
| glm | Europe (Germany) | 596 ms | 200 ms | -66% |
| replicate | South America (São Paulo) | 186 ms | 308 ms | +65% |
| reka | South America (São Paulo) | 369 ms | 203 ms | -45% |
| inference-net | Europe (Germany) | 199 ms | 111 ms | -44% |
| inference-net | US (Central) | 187 ms | 124 ms | -34% |
| hyperbolic | US (Central) | 140 ms | 180 ms | +28% |
| glm | US (Central) | 374 ms | 268 ms | -28% |
Same series: ← 17 August 2026
This post is a snapshot. The live figures behind it: /fastest-ai-api · all posts