AI Thinking Fields#

Based on a production review, understand the thinking/reasoning fields exposed by the different protocols in Crazyrouter

Updated: 2026-06-06

AI Thinking Fields#

Currently verified fields#

Note: In this review round, message.reasoning_content in gpt-5.5 Chat Completions did not return reliable content consistently, so it is not listed as a primary observation field here.

GPT: Responses reasoning item#

{
  "output": [
    {
      "id": "rs_xxx",
      "type": "reasoning",
      "encrypted_content": "...",
      "summary": [
        {
          "type": "summary_text",
          "text": "..."
        }
      ]
    },
    {
      "type": "message",
      "content": [
        {
          "type": "output_text",
          "text": "Final answer"
        }
      ]
    }
  ]
}

Claude: thinking block#

{
  "content": [
    {
      "type": "thinking",
      "thinking": "..."
    },
    {
      "type": "text",
      "text": "Final answer"
    }
  ]
}

Gemini: thoughtsTokenCount#

{
  "usageMetadata": {
    "thoughtsTokenCount": 120
  }
}

Things to keep in mind#

Last verified/modified: 2026-09-28 (reviewed; all three models gpt-5.5 / claude-opus-4-8 / gemini-3.1-pro are in the catalog)