Kaggle BrowseComp: leaderboard
OpenAI's BrowseComp (1,266 hard-to-find web questions with short verifiable answers), run by Kaggle on its own harness for every listed model, including older non-agentic ones. Kaggle's score is shown as a percentage (0-100, higher is better); results sit far below vendor-reported agent scores, so this board is not comparable to self-reported BrowseComp numbers.
Metric: Score (%). Source: www.kaggle.com. Saturation forecast: Around February 2027. 40 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 31.19 | #64 |
| 2 | GPT-5 | 20.06 | #105 |
| 3 | Gemini 3 Flash (Preview) | 16.92 | #90 |
| 4 | Gemini 2.5 Pro (Preview 06-05) | 7.82 | #118 |
| 5 | O1 | 6.32 | #190 |
| 6 | GPT-4.1 | 3.55 | #258 |
| 7 | O4 Mini | 3.32 | #185 |
| 8 | DeepSeek R1 0528 | 3.08 | #232 |
| 9 | Gemini 2.5 Flash (Preview 05-20) | 2.53 | #229 |
| 10 | Grok 3 Mini | 1.74 | #289 |
| 11 | GPT-OSS-120B | 1.58 | #346 |
| 12 | O3 Mini | 1.5 | #282 |
| 13 | DeepSeek V3 | 1.26 | #329 |
| 14 | Grok 3 | 1.26 | #313 |
| 15 | Gemini 3.1 Flash Lite (Preview) | 1.11 | #201 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=kaggle-browsecomp · How It Works · Data refreshed daily, snapshot 2026-11-05.