Kaggle DeepSearchQA (Google): leaderboard
Google DeepMind benchmark for deep research agents on 900 multi-step web information-seeking tasks requiring exhaustive answer sets, entity resolution, and stopping-criteria judgment.
Metric: F1 Score (%). Source: www.kaggle.com. Saturation forecast: Around January 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Pro | 78.98 | #81 |
| 2 | GPT-5.4 | 78.63 | #79 |
| 3 | Gemini 3.1 Pro (Preview) | 75.77 | #58 |
| 4 | GPT-5 | 73.24 | #98 |
| 5 | O3 Deep Research | 66.45 | #115 |
| 6 | O4 Mini Deep Research | 61.76 | #224 |
| 7 | Gemini 2.5 Pro | 48 | #152 |
| 8 | Gemini 2.5 Flash | 42.99 | #245 |
| 9 | Claude Opus 4.5 (Thinking) | 40.2 | #83 (Claude Opus 4.5) |
| 10 | Claude Sonnet 4.5 (Thinking) | 27.85 | #142 (Claude Sonnet 4.5) |
| 11 | Claude Haiku 4.5 (Thinking) | 22.24 | #280 (Claude Haiku 4.5) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=kaggle-deepsearchqa-google · How It Works · Data refreshed daily, snapshot 2026-10-20.