Realtime models are not ordinary chat completions. They use WebSocket sessions, so choose a model with realtime or audio support and follow its event format.
Do not infer realtime support from a model name. Read /v1/models and choose only models that explicitly expose realtime or audio input/output capabilities for your account.
curl https://api.sealink.io/v1/models \-H "Authorization: Bearer $SEALINK_API_KEY"
# 1. Choose a model that lists realtime or audio input/output in /v1/models.# 2. Use the realtime endpoint shown in /api/openapi for that model family.# 3. Verify the connection, then send a short audio sample.# 4. Follow the event format and limits shown for the selected model.
A WebSocket upgrade alone is not a billable success. Realtime charges are based on model service usage from the completed session.
