Skip to main content
Issue: Due to a recent POP adjustment (now routed through Hong Kong to improve European access performance), some instances may temporarily be unable to directly reach the official OpenAI (Codex) and Anthropic (Claude) API endpoints.

​
Temporary Solution

Please replace the original API domain with: http://llm.gpuhub.com. All request paths and headers should remain unchanged.
Use http, NOT https.

​
Domain Replacement Rules

Original OpenAI endpoint: https://api.openai.com → Replace with: http://llm.gpuhub.com Original Anthropic endpoint: https://api.anthropic.com → Replace with: http://llm.gpuhub.com

​
Quick Connectivity Test

curl http://llm.gpuhub.com/v1/models \
-H "Authorization: Bearer YOUR_API_KEY"

​
Example – Codex (OpenAI)

Replace sk-proj-XXXXXXXXXXXXXXX with your own key 
curl -i http://llm.gpuhub.com/v1/chat/completions \
-H "Authorization: Bearer sk-proj-XXXXXXXXXXXXXXX" \
-H "Content-Type: application/json" \
-d '{
  "model": "gpt-4o",
  "messages": [{"role": "user", "content": "hi"}]
}'

​
Example – Claude (Anthropic)

Replace sk-ant-XXXXXXXXXXXXXXXXXXX with your own key 
curl -i http://llm.gpuhub.com/v1/messages \
-H "x-api-key: sk-ant-XXXXXXXXXXXXXXXXXXX" \
-H "Content-Type: application/json" \
-d '{
  "model": "claude-3-5-sonnet-20240620",
  "max_tokens": 20,
  "messages": [{"role": "user", "content": "Hello Claude"}]
}'
If you see “Welcome to nginx!”, please verify that you are using http (not https) and that the request path (e.g., /v1/chat/completions) is correct.