AutoLogi: leaderboard
1,575 English and 883 Chinese open-ended logic puzzles generated automatically from AR-LSAT and LogiQA passages by Qwen Team and Fudan (2025); JSON answers checked by program verifiers, accuracy.
Metric: Score (self-reported). Source: benchmarklist.com. Status: saturated. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o (2024-08-06) | 62.04 |
| 2 | Claude 3.5 Sonnet | 60.97 |
| 3 | Llama 3.1 405B Instruct | 59.42 |
| 4 | Llama 3.1 70B Instruct | 53.79 |
| 5 | Qwen 2.5 72B Instruct | 53.5 |
| 6 | Qwen 2.5 7B Instruct | 27.28 |
| 7 | Llama 3.1 8B Instruct | 25.53 |
| 8 | GPT-3.5 Turbo | 18.93 |
Interactive version: theaggregate.ai/benchmark?slug=autologi · How It Works · Data refreshed daily, snapshot 2026-09-05.