Where is each AI API reachable from?

One request to every catalogued AI API host from eight Cloudflare regions, every 15 minutes, carrying a placeholder token that belongs to no account. Colour is the answer; status codes never reach the page.

In short: green means the API application answered from that region and would work with a real key, amber means the API refuses that region by policy, red means nothing usable came back for two rounds, grey means the probe itself was blocked so nothing can be said. Regions are Cloudflare egress points, not countries; mainland China is not covered. To see your own path, open a provider page and test from this browser.

Loading the latest round…

One provider from eight regions

Reachability matrix

Each cell is the latest probe from that region; the number is time to response headers. Hover a cell for the region, the state and the round time.

A cell turns red only after two consecutive failed rounds, and grey when the probe was blocked before the API or the region has not reported for two rounds. A green cell that took more than three seconds is marked slow: the API works but is degraded. None of this says anything about your key, quota, model or your own network path.

Each provider’s codes, streaming behaviour and a test from your own browser

Checking…

OpenAI API

api.openai.com from eight regions, and the checks that separate an OpenAI outage from your own network.

Probe OpenAI →
Checking…

Anthropic API

A live probe of api.anthropic.com from NetOkay’s edge, and how to tell a Claude API incident from an idle timeout on your own path.

Probe Anthropic →
Checking…

Google Gemini API

A live probe of generativelanguage.googleapis.com from NetOkay’s edge, and how Gemini streaming differs from the OpenAI-style SSE most clients expect.

Probe Gemini →
Checking…

Groq API

A live probe of api.groq.com from NetOkay’s edge, and why a fast provider makes your own network’s behaviour easier to spot.

Probe Groq →
Checking…

Mistral API

A live probe of api.mistral.ai from NetOkay’s edge, and the checks that separate a Mistral incident from a problem between you and the EU.

Probe Mistral →
Checking…

DeepSeek API

A live probe of api.deepseek.com from NetOkay’s edge, and why DeepSeek’s reasoning models are the most common way to discover an idle timeout on your own path.

Probe DeepSeek →
Checking…

OpenRouter API

A live probe of openrouter.ai from NetOkay’s edge, and how to tell an OpenRouter problem from an upstream model provider or your own network.

Probe OpenRouter →
Checking…

Together AI API

A live probe of api.together.xyz from NetOkay’s edge, and why a model that has not been used recently can look like a network problem.

Probe Together AI →