Qwen2-7B-Instruct is an instruction-tuned language model with 7 billion parameters, supporting a context length of up to 131,072 tokens.
发布日期2024年7月23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 |
|---|---|---|---|
| MT-Bench | communication reasoning general roleplay | 84.1 | 来源 |
| GSM8k | math reasoning | 82.3 | 来源 |
| HumanEval | reasoning code | 79.9 | 来源 |
| C-Eval | general reasoning | 77.2 | 来源 |
| AlignBench | general language math reasoning roleplay | 72.1 | 来源 |
| MMLU | general reasoning language math | 70.5 | 来源 |
| EvalPlus | reasoning code | 70.3 | 来源 |
| MBPP | reasoning general | 67.2 | 来源 |
| MultiPL-E | general language | 59.1 | 来源 |
| MATH | math reasoning | 49.6 | 来源 |
| MMLU-Pro | language reasoning math general | 44.1 | 来源 |
| LiveCodeBench | reasoning general code | 26.6 | 来源 |
| GPQA | reasoning general | 25.3 | 来源 |
| TheoremQA | math reasoning physics finance | 25.3 | 来源 |
Pricing
API 价格对比
暂无 API 价格。