Check whether an endpoint behaves like OpenAI.

Send one tiny chat-completions request from your browser and inspect the response shape, usage tokens, returned model field, and latency. It is a compatibility smoke test, not a model identity proof.

Run smoke testCheck API cost pattern

Test an OpenAI-compatible endpoint

This browser test sends one small chat-completions request directly to the endpoint you enter. Your key is not sent to TokenAir or stored by this page.

Use the OpenAI-compatible base URL, usually ending in /v1.

Optional for demo endpoints. The key stays in this browser tab.

Server-side fallback

curl 'https://api.example.com/v1/chat/completions' \
  -H 'Content-Type: application/json' \
  -d '{"model":"gpt-4o-mini","messages":[{"role":"user","content":"Reply with exactly this word and no punctuation: tokenair"}],"temperature":0,"max_tokens":16}'

Compatibility is only the first gate.

A good migration test also checks streaming, tool calling, structured output, billing visibility, retries, fallback behavior, and quality on a real workflow.

Join TokenAir early accessRead migration checklist