HumanLikeness - Syntax-1: leaderboard
HumanLikeness syntax task comparing model responses with human structural priming behavior.
Metric: Humanlike Score (%). Source: huggingface.co. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Phi-3 Mini 4K Instruct | 89.36 | #1297 |
| 2 | starchat2-15B-v0.1 | 86.93 | #832 |
| 3 | zephyr-7B-alpha | 84.56 | #1312 |
| 4 | Mistral Nemo Instruct (2407) | 84.21 | #936 |
| 5 | Llama 3.1 8B Instruct | 83.58 | #1035 |
| 6 | CodeLlama-34B-Instruct-hf | 82.34 | #1381 |
| 7 | Llama 3.1 70B Instruct | 81.26 | #561 |
| 8 | Llama 3 8B Instruct | 78.94 | #1129 |
| 9 | GPT-3.5 Turbo | 76.33 | #865 |
| 10 | Llama 2 13B Chat Base | 74.98 | #1422 |
| 11 | Mistral 7B Instruct (v0.3) | 74.89 | #1274 |
| 12 | c4ai-command-r-plus | 72.36 | #724 |
| 13 | Mistral 7B Instruct (v0.2) | 72.27 | #1333 |
| 14 | Yi 1.5 34B Chat | 72.24 | #800 |
| 15 | zephyr-7B-beta | 71.37 | #1330 |
Interactive version: theaggregate.ai/benchmark?slug=humanlikeness-syntax-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.