加载中...
加载中...
BIG-Bench 的困难子集，包含更具挑战性的任务，用于评估模型的极限能力。
Source: DataLearnerAI
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
| 排名 | 模型 | |||
|---|---|---|---|---|
| 1 | ERNIE-4.5-300B-A47Bdefault | 94.3 | 2025-06-30 | 3000 |
| 2 | Claude 3.5 Sonnet Newdefault | 92.6 | 2024-10-22 | 未知 |
| 3 | DeepSeek-V3default | 92.3 | 2024-12-26 | 6810 |
| 4 | Hunyuan-TurboSdefault | 92.2 | 2025-03-10 | 未知 |
| 5 | GPT-4odefault | 91.7 | 2024-05-13 | 未知 |
| 6 | Llama3.1-405B Instructdefault | 89.2 | 2024-07-23 | 4050 |
| 7 | Hunyuan-A13B-Instructdefault | 89.1 | 2025-06-27 | 800 |
| 8 | Qwen3-235B-A22Bdefault | 88.87 | 2025-04-28 | 2350 |
| 9 | Qwen3-Nextdefault | 87.13 | 2025-09-11 | 800 |
| 10 | Qwen2.5-72Bdefault | 86.3 | 2024-09-18 | 727 |
| 11 | Gemma 2 - 9Bdefault | 68.2 | 2024-06-27 | 90 |
| 12 | Moonlight-16B-A3B-Instructdefault | 65.2 | 2025-02-23 | 160 |
| 13 | Llama3.1-8Bdefault | 57.7 | 2024-07-23 | 80 |
| 14 | Qwen2.5-3Bdefault | 56.3 | 2024-09-18 | 30 |
| 15 | Mistral-7B-Instruct-v0.3default | 56.1 | 2024-05-22 | 70 |
| 16 | Llama-3.2-3Bdefault | 46.8 | 2024-09-18 | 32 |
| 17 | Gemini 1.5 Prodefault | 0 | 2024-02-15 | 未知 |
| 18 | Amazon Nova Prodefault | 0 | 2024-12-03 | 未知 |