Google Gemini 3 Deep Think - Humanity's Last Exam (no tools): leaderboard
Metric: Score (%). Source: storage.googleapis.com. 4 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Deep Think | 48.4 |
| 2 | Qwen 3.7 Max | 41.4 |
| 3 | Claude Opus 4.6 (Thinking) | 40 |
| 4 | DeepSeek V4 Pro (Reasoning, Max Effort) | 37.7 |
| 5 | Gemini 3 Pro (Preview) | 37.5 |
| 6 | Kimi K2.6 (Thinking) | 36.4 |
| 7 | GLM-5.1 (Thinking) | 34.7 |
| 8 | GPT-5.2 (xHigh) | 34.5 |
| 9 | Qwen 3.6 Plus | 28.8 |
Interactive version: theaggregate.ai/benchmark?slug=google-gemini-3-deep-think-humanity-s-last-exam-no-tools · How It Works · Data refreshed daily, snapshot 2026-09-05.