Embeddable AI API status & latency badges
Is the OpenAI, Claude or Gemini API up? Show it in your README
A status badge reads up, degraded or down from the same probes that run every five minutes from 4 regions; the image is rebuilt about every 30 minutes and links to the provider’s live status page. Down means two or more regions cannot reach the API; degraded means one region failed or the provider’s own status page has an open incident.
[](https://llmlatency.dev/provider/openai)Swap openai for any provider key below (anthropic, google,
mistral, …). Same licence, no sign-up, no tracking.
Latency badge
A small SVG badge showing an API’s independently measured edge latency. It updates itself: the image is regenerated from live measurements, so whatever you embed stays current without you touching it.
How do I add the badge?
Markdown — works in a GitHub README:
[](https://llmlatency.dev/provider/fireworks)HTML:
<a href="https://llmlatency.dev/provider/fireworks"><img src="https://llmlatency.dev/badge/fireworks.svg" alt="fireworks latency"></a>Replace the provider name in the URL with any provider we measure. The badge shows that provider’s fastest measured region and names it, so the number is never quoted without its origin.
Which badges are available?
| Provider | Badge | p50 TTFB | Fastest region |
|---|---|---|---|
fireworks |  | 22 ms | Asia (Tokyo) |
sambanova |  | 24 ms | Asia (Tokyo) |
baseten |  | 47 ms | US (Central) |
google |  | 51 ms | US (Central) |
meta-llama |  | 55 ms | US (Central) |
openrouter |  | 58 ms | Asia (Tokyo) |
cohere |  | 58 ms | US (Central) |
upstage |  | 59 ms | Asia (Tokyo) |
typesafe |  | 75 ms | US (Central) |
novita |  | 78 ms | US (Central) |
cerebras |  | 80 ms | US (Central) |
siliconflow |  | 84 ms | US (Central) |
replicate |  | 86 ms | US (Central) |
perplexity |  | 88 ms | US (Central) |
anthropic |  | 90 ms | US (Central) |
targon |  | 91 ms | US (Central) |
glm |  | 92 ms | Asia (Tokyo) |
reka |  | 94 ms | US (Central) |
nscale |  | 98 ms | Europe (Germany) |
nebius |  | 100 ms | Europe (Germany) |
What exactly does the number mean?
Edge latency (time-to-first-byte) at the 50th percentile over the last 24 hours, measured by probes that connect directly to the provider’s own API host — never through a gateway. The badge names the region the figure comes from, because a latency number without an origin is not a fact about a provider. Full method on the methodology page.
Can I use this commercially?
Yes. The badge and the underlying data are licensed CC BY 4.0 — free to use, including commercially, as long as llmlatency.dev is credited. We do not track who embeds it and there is nothing to sign up for. Providers are welcome to display their own figure; the number shown is the same one published in our public tables and is not adjustable by anyone, including us.