SeaEval - Fundamental NLP Tasks - MRPC (Zero-Shot): leaderboard
SeaEval zero-shot MRPC evaluation measuring core NLP understanding such as inference, acceptability, paraphrase, or reading-comprehension accuracy.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2 72B Instruct | 80.64 | #576 |
| 2 | Qwen 2.5 72B Instruct | 80.15 | #454 |
| 3 | Qwen 2 7B Instruct | 78.68 | #991 |
| 4 | Qwen 2.5 14B Instruct | 77.94 | #647 |
| 5 | gemma2-9B-cpt-sea-lionv3-instruct | 77.94 | #705 |
| 6 | Sailor2-8B-Chat | 77.7 | #782 |
| 7 | Qwen 2.5 32B Instruct | 77.45 | #498 |
| 8 | Llama 3.1 70B Cpt Sea Lionv3 Instruct | 77.45 | #548 |
| 9 | Llama 3.1 70B Instruct | 76.96 | #561 |
| 10 | Llama 3 70B Instruct | 75.98 | #634 |
| 11 | SeaLLMs-v3-7B-Chat | 74.75 | #915 |
| 12 | Gemma 2 9B (IT) | 74.02 | #788 |
| 13 | Gemma 2 2B (IT) | 70.83 | #1341 |
| 14 | Qwen 2.5 7B Instruct | 70.59 | #862 |
| 15 | Qwen 2.5 1.5B Instruct | 68.38 | #1407 |
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-mrpc-zero-shot · How It Works · Data refreshed daily, snapshot 2026-10-20.