LLM2014 Logic 2025-12: leaderboard
Metric: Median Score. Source: raw.githubusercontent.com. 51 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 (High) | 81.83 |
| 2 | Gemini 3 Pro | 79.29 |
| 3 | GPT-5.1 (High) | 77.54 |
| 4 | Gemini 3 Flash (High) | 76.9 |
| 5 | GPT-5 Mini (High) | 75.05 |
| 6 | DeepSeek V3.2 Speciale | 72.68 |
| 7 | Grok 4 | 65.61 |
| 8 | GLM-4.7 (Thinking) | 65.49 |
| 9 | Claude Sonnet 4.5 (Thinking) | 62.7 |
| 10 | Kimi K2 (Thinking) | 59.34 |
| 11 | Grok 4.1 Fast (Reasoning) | 57.68 |
| 12 | Qwen 3 Max (Thinking) | 49.68 |
| 13 | Qwen 3 235B A22B 2507 (Thinking) | 46.71 |
| 14 | Qwen 3 Max | 42.85 |
| 15 | GPT-OSS-120B (High) | 41.96 |
Interactive version: theaggregate.ai/benchmark?slug=llm2014-logic-2025-12 · How It Works · Data refreshed daily, snapshot 2026-09-05.