HumanLikeness - Syntax-2: leaderboard
HumanLikeness syntax task comparing model responses with human syntactic ambiguity resolution.
Metric: Humanlike Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3 8B Instruct | 79.49 | #1129 |
| 2 | Llama 3.1 8B Instruct | 79.21 | #1035 |
| 3 | Llama 3.1 70B Instruct | 78.42 | #561 |
| 4 | GPT-4o | 78.35 | #343 |
| 5 | Mistral Nemo Instruct (2407) | 77.18 | #936 |
| 6 | GPT-3.5 Turbo | 77.07 | #865 |
| 7 | GPT-4o Mini | 76.94 | #599 |
| 8 | Llama 3 70B Instruct | 76.72 | #634 |
| 9 | starchat2-15B-v0.1 | 73.32 | #832 |
| 10 | c4ai-command-r-plus | 72.4 | #724 |
| 11 | Yi 1.5 34B Chat | 70.6 | #800 |
| 12 | Phi-3 Mini 4K Instruct | 69.29 | #1297 |
| 13 | Llama 2 7B Chat | 67.19 | #1499 |
| 14 | Llama 2 13B Chat Base | 65.18 | #1422 |
| 15 | Mistral 7B Instruct (v0.2) | 58.71 | #1333 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=humanlikeness-syntax-2 · How It Works · Data refreshed daily, snapshot 2026-10-20.