Qwen2.5-Coder is a specialized coding model trained on 5.5 trillion tokens of code data, supporting 92 programming languages with a 128K context window. It excels in code generation, completion, repair, and multi-programming tasks while maintaining strong performance in mathematics and general capabilities.
发布日期2024年9月19日
参数规模32B
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 |
|---|---|---|---|
| HumanEval | reasoning code | 92.7 | 来源 |
| GSM8k | math reasoning | 91.1 | 来源 |
| MBPP | reasoning general | 90.2 | 来源 |
| HellaSwag | reasoning | 83.0 | 来源 |
| Winogrande | reasoning language | 80.8 | 来源 |
| MMLU-Redux | language reasoning math general | 77.5 | 来源 |
| MMLU | general reasoning language math | 75.1 | 来源 |
| ARC-C | reasoning general | 70.5 | 来源 |
| MATH | math reasoning | 57.2 | 来源 |
| TruthfulQA | general reasoning legal healthcare finance | 54.2 | 来源 |
| MMLU-Pro | language reasoning math general | 50.4 | 来源 |
| BigCodeBench-Full | general reasoning | 49.6 | 来源 |
| TheoremQA | math reasoning physics finance | 43.1 | 来源 |
| LiveCodeBench | reasoning general code | 31.4 | 来源 |
| BigCodeBench-Hard | general reasoning | 27.0 | 来源 |
Pricing
API 价格对比
暂无 API 价格。