ChatGPT Model TrackerLatest reading Oct 22, 15:00 UTC

Which model answers ChatGPT right now

Every day querying.ai sends the same set of questions to ChatGPT in logged-out sessions and records which model answers and what each answer carries.

Default model now

gpt-6

100% of the last 24 hours' answers

Default since
Oct 9, 07:00 UTC
14 days ago
Previous default
gpt-5-6
led for 35 days

Model by day

  • gpt-5-5
  • gpt-5-6
  • gpt-6
gpt-5-6Sep 4 – Oct 9gpt-6Oct 9 – now

Recorded answers before and after gpt-6

Compare recorded components for the same daily questions. Historical widget citations were partially omitted; these rates reflect both model behavior and parser coverage.

Share of answers

Averages

Inline citations

  • gpt-5-6avg 91.2%
  • gpt-6avg 58.2%

Answers with source chips inside the text, next to the claim each one supports.

Use the data

The tracker's data is open as JSON. No key needed.

GEThttps://api.querying.ai/v1/chatgpt-modelOpen JSON 
{
  "success": true,
  "data": {
    "asOf": "2026-10-22T15:00:00.000Z",
    "current": { "model": "gpt-6", "share": 100, "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-6",
        "from": "2026-10-09T07:00:00.000Z",
        "to": null,
        "surfaces": {
          "entities": 79.7,
          "shopping": 3,
          "map": 1.4,
          "inlineProducts": 15,
          "citationPills": 58.2,
          "inlineSourcesPerAnswer": 2.41,
          "sourcesPerAnswer": 19.32,
          "searchesPerAnswer": 1.21
        }
      }
    ]
  }
}

How the tracker works

  1. Component rates measure what the collector and parser captured. Parser updates change coverage. Historical widget citations were undercounted; model effects require a comparison using the same parser and collection conditions.
  2. Every day querying.ai sends the same question set to ChatGPT in logged-out sessions. Only these answers feed the tracker.
  3. Each answer records the model that wrote it and what it carries: inline citations, entity cards, shopping cards, inline products and maps, plus the sources and searches behind it.
  4. A model is the default when it writes most answers for several hours in a row. An hour or a day with too few answers shows no rate.

All times in UTC