SuperCLUE General (July 2025) - Science Reasoning: leaderboard
Metric: Score. Source: www.superclueai.com. 47 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O3 (High) | 59.7 |
| 2 | O4 Mini (High) | 55.56 |
| 3 | GPT-5 (High) | 54.2 |
| 4 | Gemini 2.5 Pro | 52.59 |
| 5 | DeepSeek R1 0528 | 48.15 |
| 6 | GPT-OSS-120B | 47.41 |
| 7 | DeepSeek V3.1 (Thinking) | 47.41 |
| 8 | Claude Opus 4 (Thinking) | 45.93 |
| 9 | GLM-4.5 | 44.62 |
| 10 | Gemini 2.5 Flash | 44.44 |
| 11 | Grok 4 | 44.44 |
| 12 | Qwen 3 235B A22B 2507 (Thinking) | 42.22 |
| 13 | Qwen 3 235B A22B 2507 Instruct | 41.48 |
| 14 | Doubao-Seed-1.6-thinking-250715 | 40.49 |
| 15 | SenseNova V6 Reasoner | 40 |
Interactive version: theaggregate.ai/benchmark?slug=superclue-general-july-2025-science-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-19.