# Sample: weekly provider report

Period: 08 Sep 2026 through 14 Sep 2026 (seven complete UTC days). A working sample for five selected providers, regenerated with the site. The paid work is choosing your scope, preparing the feed and supporting delivery; this sample is free.

Generated 15 Sep 2026, 17:46 UTC

## Network edge response — TTFB

Probe failures are not confirmed provider outages. Sparse means fewer than 10 observations.

| Provider / model | Region | Probes | Failed | Success % | p50 ms | p95 ms |
|---|---|---|---|---|---|---|
| anthropic | ap-tokyo | 2019 | 0 | 100.0 | 234.0 | 306.7 |
| anthropic | eu-hetzner | 2018 | 0 | 100.0 | 199.5 | 304.0 |
| anthropic | sa-east | 2021 | 0 | 100.0 | 201.3 | 257.8 |
| anthropic | us-central | 2018 | 0 | 100.0 | 92.8 | 160.7 |
| google | ap-tokyo | 2019 | 0 | 100.0 | 61.8 | 103.8 |
| google | eu-hetzner | 2018 | 0 | 100.0 | 98.6 | 201.4 |
| google | sa-east | 2021 | 0 | 100.0 | 158.8 | 818.2 |
| google | us-central | 2018 | 0 | 100.0 | 45.9 | 106.3 |
| groq | ap-tokyo | 2019 | 0 | 100.0 | 144.7 | 198.1 |
| groq | eu-hetzner | 2018 | 0 | 100.0 | 295.6 | 402.1 |
| groq | sa-east | 2021 | 0 | 100.0 | 231.7 | 286.2 |
| groq | us-central | 2018 | 0 | 100.0 | 126.1 | 195.1 |
| mistral | ap-tokyo | 2019 | 0 | 100.0 | 313.1 | 390.2 |
| mistral | eu-hetzner | 2018 | 0 | 100.0 | 100.6 | 203.1 |
| mistral | sa-east | 2021 | 0 | 100.0 | 251.7 | 323.8 |
| mistral | us-central | 2018 | 0 | 100.0 | 191.4 | 270.9 |
| openai | ap-tokyo | 2019 | 7 | 99.653 | 203.8 | 286.7 |
| openai | eu-hetzner | 2018 | 0 | 100.0 | 206.3 | 398.4 |
| openai | sa-east | 2020 | 0 | 100.0 | 240.1 | 468.8 |
| openai | us-central | 2018 | 0 | 100.0 | 127.6 | 205.4 |

## Measured model response — TTFT

Probe failures are not confirmed provider outages. Sparse means fewer than 10 observations.

| Provider / model | Region | Probes | Failed | Success % | p50 ms | p95 ms |
|---|---|---|---|---|---|---|
| anthropic / claude-haiku-4-5 [sparse] | eu-hetzner | 1 | 1 | 0.0 | — | — |
| google / gemini-flash-lite-latest | ap-tokyo | 506 | 23 | 95.455 | 1486.6 | 3266.3 |
| google / gemini-flash-lite-latest | eu-hetzner | 503 | 27 | 94.632 | 1006.5 | 3052.5 |
| google / gemini-flash-lite-latest | sa-east | 506 | 113 | 77.668 | 1275.3 | 8046.4 |
| google / gemini-flash-lite-latest | us-central | 506 | 24 | 95.257 | 1500.1 | 4007.1 |
| groq / openai/gpt-oss-120b | ap-tokyo | 506 | 0 | 100.0 | 1374.5 | 1751.8 |
| groq / openai/gpt-oss-120b | eu-hetzner | 503 | 0 | 100.0 | 873.6 | 1297.7 |
| groq / openai/gpt-oss-120b | sa-east | 506 | 0 | 100.0 | 775.9 | 1275.4 |
| groq / openai/gpt-oss-120b | us-central | 506 | 0 | 100.0 | 1345.8 | 1777.9 |

## How to read this report

- Network TTFB measures the public API edge, including expected authentication responses; it does not measure model speed or account availability.
- Inference TTFT is measured only for the provider/model pairs shown. Different models are not a like-for-like provider benchmark.
- Failed probes can reflect test credentials, account quotas or network problems; they are not confirmed provider outages. Series with fewer than 10 probes are marked sparse and do not establish ongoing coverage.
- Success is the fraction of received probes marked ok, not time-based uptime or an SLA. Missing probes are not inferred to be successful or failed.
- Percentiles use successful non-null samples across the whole seven-day window, not averages of daily percentiles. Sorted index = floor((p/100)*(n-1)+0.5).
- Unobserved series are omitted; null latency means no successful timed sample. This sample is a format preview, not a customer deployment.

[JSON](https://llmlatency.dev/api/pilot-sample.json) · [Source](https://llmlatency.dev/data) · [Methodology](https://llmlatency.dev/methodology)

CC BY 4.0.