Qwen2-72B-Instruct is an instruction-tuned language model with 72 billion parameters, supporting a context length of up to 131,072 tokens. It's part of the new Qwen2 series, which has surpassed most open-source models and demonstrates competitiveness against proprietary models across various benchmarks.
发布日期2024年7月23日
参数规模72B
上下文长度—
许可证tongyi-qianwen
知识截止—
Benchmarks
评测成绩
| 评测基准 | 类别 | 分数 | 来源 |
|---|---|---|---|
| GSM8k | math reasoning | 91.1 | 来源 |
| CMMLU | language reasoning general | 90.1 | 来源 |
| HellaSwag | reasoning | 87.6 | 来源 |
| HumanEval | reasoning code | 86.0 | 来源 |
| Winogrande | reasoning language | 85.1 | 来源 |
| C-Eval | general reasoning | 83.8 | 来源 |
| BBH | reasoning math language | 82.4 | 来源 |
| MMLU | general reasoning language math | 82.3 | 来源 |
| MBPP | reasoning general | 80.2 | 来源 |
| EvalPlus | reasoning code | 79.0 | 来源 |
| MultiPL-E | general language | 69.2 | 来源 |
| ARC-C | reasoning general | 68.9 | 来源 |
| MMLU-Pro | language reasoning math general | 64.4 | 来源 |
| MATH | math reasoning | 59.7 | 来源 |
| TruthfulQA | general reasoning legal healthcare finance | 54.8 | 来源 |
| TheoremQA | math reasoning physics finance | 44.4 | 来源 |
| GPQA | reasoning general | 42.4 | 来源 |
Pricing
API 价格对比
暂无 API 价格。