HumanLikeness - Word-1: leaderboard
HumanLikeness word-level task comparing model responses with human word length and predictability judgments.
Metric: Humanlike Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 70.88 | #343 |
| 2 | GPT-3.5 Turbo | 68.9 | #865 |
| 3 | c4ai-command-r-plus | 65.8 | #724 |
| 4 | GPT-4o Mini | 64.81 | #599 |
| 5 | Llama 3.1 70B Instruct | 63.71 | #561 |
| 6 | CodeLlama-34B-Instruct-hf | 62.43 | #1381 |
| 7 | Llama 3.1 8B Instruct | 62.18 | #1035 |
| 8 | Phi-3 Mini 4K Instruct | 61.24 | #1297 |
| 9 | Llama 2 7B Chat | 60.24 | #1499 |
| 10 | starchat2-15B-v0.1 | 60.05 | #832 |
| 11 | Llama 3 8B Instruct | 59.96 | #1129 |
| 12 | Mistral Nemo Instruct (2407) | 58.69 | #936 |
| 13 | Llama 3 70B Instruct | 57.26 | #634 |
| 14 | Yi 1.5 34B Chat | 57.21 | #800 |
| 15 | Mistral 7B Instruct (v0.2) | 56.32 | #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-word-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.