SuperCLUE General (March 2026) - Science Reasoning: leaderboard
Metric: Score. Source: www.superclueai.com. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 (Max) | 85.37 |
| 2 | GPT-5.4 (xHigh) | 84.15 |
| 3 | Gemini 3.1 Pro (Preview) (High) | 81.71 |
| 4 | Gemini 3 Flash (Preview) (High) | 81.71 |
| 5 | DeepSeek V4 Pro (Max) | 79.27 |
| 6 | DeepSeek V4 Flash (Max) | 79.01 |
| 7 | Grok 4.20 Beta (0309) (Reasoning) | 78.05 |
| 8 | Qwen 3.5 397B A17B (Thinking) | 75.61 |
| 9 | GLM-5 | 75 |
| 10 | MiMo-V2-Pro | 74.39 |
| 11 | GPT-OSS-120B (High) | 73.17 |
| 12 | DeepSeek V3.2 (Thinking) | 73.17 |
| 13 | Spark X2 | 71.95 |
| 14 | Tencent HY 2.0 Think | 70.73 |
| 15 | Qwen 3.5 122B A10B (Thinking) | 69.51 |
Interactive version: theaggregate.ai/benchmark?slug=superclue-general-march-2026-science-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-19.