Qwen2.5-7B-Instruct is an instruction-tuned 7B parameter language model that excels at following instructions, generating long texts (over 8K tokens), understanding structured data, and generating structured outputs like JSON. The model features enhanced capabilities in mathematics, coding, and multilingual support across 29+ languages including Chinese, English, French, Spanish, and more.
发布日期2024年9月19日
参数规模7.6B
上下文长度—
许可证Apache 2.0
知识截止—
Benchmarks
评测成绩
| 评测基准 | 类别 | 分数 | 来源 |
|---|---|---|---|
| GSM8k | math reasoning | 91.6 | 来源 |
| 安全防御榜·Toxicity | security | 90.2 | — |
| MT-Bench | communication reasoning general roleplay | 87.5 | 来源 |
| 安全防御榜·Bias | security | 86.8 | — |
| HumanEval | reasoning code | 84.8 | 来源 |
| 安全防御榜·Info Safety | security | 84.3 | — |
| 安全防御榜·Human Rights | security | 83.1 | — |
| 安全防御榜 | security | 82.2 | — |
| 安全防御榜·Malicious Use | security | 80.4 | — |
| 价值对齐榜·Fairness | security | 79.4 | — |
| MBPP | reasoning general | 79.2 | 来源 |
| MATH | math reasoning | 75.5 | 来源 |
| MMLU-Redux | language reasoning math general | 75.4 | 来源 |
| 安全防御榜·Child Safety | security | 75.4 | — |
| 安全防御榜·Socioeconomic | security | 75.0 | — |
| AlignBench | general language math reasoning roleplay | 73.3 | 来源 |
| IFEval | general | 71.2 | 来源 |
| MultiPL-E | general language | 70.4 | 来源 |
| 价值对齐榜·Morality | security | 62.9 | — |
| 价值对齐榜 | security | 58.6 | — |
| 价值对齐榜·Safety | security | 57.1 | — |
| 价值对齐榜·Legality | security | 56.7 | — |
| MMLU-Pro | language reasoning math general | 56.3 | 来源 |
| Arena Hard | general reasoning creativity | 52.0 | 来源 |
| 价值对齐榜·Data Protection | security | 36.8 | — |
| GPQA | reasoning general | 36.4 | 来源 |
| LiveBench | math reasoning general | 35.9 | 来源 |
| LiveCodeBench | reasoning general code | 28.7 | 来源 |
Pricing
API 价格对比
暂无 API 价格。