Kaggle FACTS (Google): leaderboard
Four factuality sets in one number: trivia questions answered from the model's own knowledge, questions answered with a shared web search tool, questions about an input image, and answers that must stay grounded in the context supplied in the prompt. Kaggle owns a held-out private set for each and runs the models itself. This board is the suite's average accuracy over public and private sets, higher is better.
Metric: Avg Score (%). Source: www.kaggle.com. Saturation forecast: Around June 2027. 39 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 75.68 | #6 |
| 2 | GPT-6.1 Sol | 75.29 | #7 |
| 3 | GPT-6 Sol | 72.5 | #14 |
| 4 | Gemini 3.7 Flash | 71.25 | #23 |
| 5 | Gemini 3.8 Flash | 69.37 | #21 |
| 6 | Gemini 3.1 Pro (Preview) | 67.42 | #58 |
| 7 | GPT-5.5 | 66.67 | #31 |
| 8 | Gemini 3.6 Flash | 65.31 | #51 |
| 9 | Claude Opus 5.5 | 65.3 | #3 |
| 10 | Gemini 3 Flash (Preview) | 63.36 | #82 |
| 11 | GPT-6 Luna | 63.28 | #56 |
| 12 | Gemini 3.5 Flash | 63.11 | #60 |
| 13 | Gemini 2.5 Pro | 61.45 | #152 |
| 14 | GPT-5 | 56.5 | #98 |
| 15 | Claude Opus 4.8 | 53.53 | #37 |
No result here: #8 Gemini 4 Argon, #9 Claude Fable 5.1, #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=kaggle-facts-google · How It Works · Data refreshed daily, snapshot 2026-10-20.