Introducing the ChatGPT Model Tracker

See which model answers logged-out ChatGPT, when the default changed and what each model’s answers carry, on a public page and an open API.

What you get

• A public page at querying.ai/en/chatgpt-model in eight languages: the default model now, the last switch and the date tracking began.
• The daily share of answers by model since September 4, 2026, with each switch marked on the chart.
• For each model, the share of its answers that carry entity cards, shopping cards, inline products and maps.
• The same data as JSON from GET /v1/chatgpt-model, open to anyone.

How it works

Every day querying.ai sends its own fixed question set to ChatGPT in logged-out sessions and records the model behind each answer and the components it carries. The tracker updates every hour, and its figures describe the default model of logged-out ChatGPT.

A model becomes the default when it writes most answers for 3 qualifying hours in a row, and that hour is recorded as a switch. A rate appears for an hour with at least 30 answers and for a day with at least 200. Component rates follow the questions asked, so the page compares each model over the stretch it led, leaving out the first 24 hours after a switch.

On October 9, 2026 at 07:00 UTC the default moved from gpt-5-6 to gpt-6. Answers with shopping cards averaged 24.8% under gpt-5-6 and 3.6% under gpt-6.

Where it helps

• Check whether ChatGPT switched models on the day your brand’s mentions or citations moved.
• Mark model switches on your own monitor charts and client reports.
• Read the API once a day and alert your team when current.model changes.
• Follow how often ChatGPT answers carry shopping cards and maps when you plan product and local content.

Get started

Call the API without a key. Responses are cached for 5 minutes, and each IP can send 240 requests per minute. Above that, the API answers 429 with RATE_LIMITED.

curl https://api.querying.ai/v1/chatgpt-model

Response

{
  "success": true,
  "data": {
    "generatedAt": "…",
    "asOf": "…",
    "models": ["gpt-5-5", "gpt-5-6", "gpt-6"],
    "current": {
      "model": "gpt-6",
      "share": 99.8,
      "since": "2026-10-09T07:00:00.000Z"
    },
    "switches": [
      { "at": "2026-10-09T07:00:00.000Z", "from": "gpt-5-6", "to": "gpt-6" }
    ],
    "eras": [
      {
        "model": "gpt-5-6",
        "from": "2026-09-04T13:00:00.000Z",
        "to": "2026-10-09T07:00:00.000Z",
        "surfaces": {
          "entities": 63.7,
          "shopping": 24.8,
          "map": 1.7,
          "inlineProducts": 20.4,
          "citationPills": 91.2,
          "inlineSourcesPerAnswer": 2.93,
          "sourcesPerAnswer": 16.74,
          "searchesPerAnswer": 1.35
        }
      },
      {
        "model": "gpt-6",
        "from": "2026-10-09T07:00:00.000Z",
        "to": null,
        "surfaces": {
          "entities": 81.7,
          "shopping": 3.6,
          "map": 0.6,
          "inlineProducts": 3.8,
          "citationPills": 22.1,
          "inlineSourcesPerAnswer": 0.69,
          "sourcesPerAnswer": 18.41,
          "searchesPerAnswer": 1.19
        }
      }
    ],
    "daily": [
      "…",
      {
        "day": "2026-10-11",
        "shares": { "gpt-6": 99.8, "gpt-5-6": 0.2 },
        "surfaces": {
          "entities": 81.6,
          "shopping": 3.9,
          "map": 0.2,
          "inlineProducts": 4.3,
          "citationPills": 20.2,
          "inlineSourcesPerAnswer": 0.62,
          "sourcesPerAnswer": 18.48,
          "searchesPerAnswer": 1.19
        }
      }
    ]
  }
}

currentobject

The default model now, its share of the last 24 hours’ answers (share, percent) and when it took over (since).

switchesarray

Every switch of the default model, with at, from and to.

erasarray

One entry per stretch a model led, with from, to (null while it still leads) and surfaces.

dailyarray

Each day with enough answers, oldest first: shares per model in percent, and surfaces.

surfacesobject

The share of answers that carry entity cards (entities), shopping cards (shopping), inline products (inlineProducts), maps (map) and in-text citations (citationPills), in percent. inlineSourcesPerAnswer is the average number of distinct sources an answer cites in its text, sourcesPerAnswer the average number of sources in its sources panel and searchesPerAnswer the average number of searches, all as counts.

modelsarray

Every model seen, in order of first appearance.

Pricing

The page and the API are free for everyone, with or without a querying.ai account.