Llama 3.2 3B Instruct is a large language model that supports a context length of 128K tokens and are state-of-the-art in their class for on-device use cases like summarization, instruction following, and rewriting tasks running locally at the edge.
发布日期2024年9月25日
参数规模3.2B
上下文长度33K
许可证Llama 3.2 Community License
知识截止—
Benchmarks
评测成绩
| 评测基准 | 类别 | 分数 | 来源 |
|---|---|---|---|
| NIH/Multi-needle | long_context | 84.7 | 来源 |
| ARC-C | reasoning general | 78.6 | 来源 |
| GSM8k | math reasoning | 77.7 | 来源 |
| IFEval | general | 77.4 | 来源 |
| HellaSwag | reasoning | 69.8 | 来源 |
| BFCL v2 | general reasoning | 67.0 | 来源 |
| MMLU | general reasoning language math | 63.4 | 来源 |
| InfiniteBench/En.MC | long_context | 63.3 | 来源 |
| MGSM | math reasoning | 58.2 | 来源 |
| MATH | math reasoning | 48.0 | 来源 |
| Open-rewrite | language writing | 40.1 | 来源 |
| Nexus | general | 34.3 | 来源 |
| GPQA | reasoning general | 32.8 | 来源 |
| InfiniteBench/En.QA | long_context | 19.8 | 来源 |
| TLDR9+ (test) | summarization language | 19.0 | 来源 |
Pricing
API 价格对比
| 服务商 | 输入价 | 输出价 | 上下文 | 吞吐(tok/s) | 延迟(s) | 函数调用 | 代码执行 | 联网搜索 |
|---|---|---|---|---|---|---|---|---|
| LLM Gateway | $0.03 | $0.05 | 33K | — | — | ✗ | ✗ | ✗ |
价格单位：美元/百万 token，数据来自社区整理，仅供参考。