Overview

GPT-5 Pro runs advanced assistants that coordinate multiple tools, interpret images, and hold realtime spoken conversations. It supports OpenAI-compatible schemas, making it easy to migrate existing workloads.

Text Mode

curl -X POST "https://api.transendai.net/v1/chat/gpt-5-pro" \
  -H "Authorization: Bearer $TRANSEND_API_KEY" \
  -H "Content-Type": "application/json" \
  -d '{
    "model": "gpt-5-pro",
    "messages": [
      { "role": "system", "content": "You are an AI operator that handles CRM updates." },
      { "role": "user", "content": "Log a Jira bug for login timeouts and ping the incident channel." }
    ],
    "tools": [
      {
        "name": "createTicket",
        "description": "Create a Jira ticket",
        "parameters": {
          "type": "object",
          "properties": {
            "title": { "type": "string" },
            "priority": { "type": "string", "enum": ["P0","P1","P2","P3"] }
          },
          "required": ["title"]
        }
      },
      {
        "name": "notifySlack",
        "description": "Send message to Slack channel",
        "parameters": {
          "type": "object",
          "properties": {
            "channel": { "type": "string" },
            "message": { "type": "string" }
          },
          "required": ["channel","message"]
        }
      }
    ]
  }'

The model decides which tool to call, returning a sequence of tool invocations and final responses.

Realtime Mode

To initiate a low-latency voice session, connect via WebSocket using the Transend Realtime Gateway:

wss://rt.transendai.net/v1/realtime?model=gpt-5-pro&api_key=sk-...

Send/receive audio frames encoded as 16-bit PCM. Responses stream back with transcripts and optional function calls.

Vision Inputs

Add image attachments by referencing URLs:

{
  "messages": [
    {
      "role": "user",
      "content": [
        { "type": "text", "text": "Summarise the chart and recommend next steps." },
        { "type": "image_url", "image_url": "https://storage.example.com/latency-chart.png" }
      ]
    }
  ]
}

Error Codes

Code	Meaning	Fix
`400`	Invalid tool schema or unsupported mode.	Verify JSON Schema and ensure `mode` is `text` or `realtime`.
`409`	Concurrent realtime session limit hit.	Close unused sessions or request higher concurrency.
`502`	Downstream provider error.	Retry with exponential backoff; the router will fail over automatically.

Best Practices

Define narrow, well-typed tool schemas so GPT-5 Pro returns actionable payloads.
Log tool_invocations to your monitoring stack for auditability.
For realtime deployments, reuse a single WebSocket per agent to avoid frequent handshakes.

GPT-5 Pro landing page
Support for realtime SLA upgrades