SeaEval - Fundamental NLP Tasks - WNLI (Zero-Shot): leaderboard
SeaEval zero-shot WNLI evaluation measuring core NLP understanding such as inference, acceptability, paraphrase, or reading-comprehension accuracy.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 25 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2 72B Instruct | 88.73 | #576 |
| 2 | Qwen 2.5 32B Instruct | 87.32 | #498 |
| 3 | Llama 3.1 70B Cpt Sea Lionv3 Instruct | 85.92 | #548 |
| 4 | Llama 3.1 70B Instruct | 84.51 | #561 |
| 5 | Qwen 2.5 14B Instruct | 83.1 | #647 |
| 6 | Qwen 2.5 72B Instruct | 81.69 | #454 |
| 7 | Llama 3 70B Instruct | 78.87 | #634 |
| 8 | Gemma 2 9B (IT) | 77.46 | #788 |
| 9 | Qwen 2.5 7B Instruct | 76.06 | #862 |
| 10 | gemma2-9B-cpt-sea-lionv3-instruct | 76.06 | #705 |
| 11 | Llama 3.1 8B CPT SEA LIONv3 Instruct | 73.24 | #810 |
| 12 | Qwen 2 7B Instruct | 71.83 | #991 |
| 13 | Qwen 2.5 3B Instruct | 64.79 | #1162 |
| 14 | Llama 3.1 8B Instruct | 61.97 | #1035 |
| 15 | SeaLLMs-v3-7B-Chat | 59.15 | #915 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=seaeval-fundamental-nlp-tasks-wnli-zero-shot · How It Works · Data refreshed daily, snapshot 2026-10-20.