Llama 3.1 70B Instruct is a large language model optimized for multilingual dialogue use cases. It outperforms many available open source and closed chat models on common industry benchmarks.
发布日期2024年7月23日
参数规模70B
上下文长度128K
许可证Llama 3.1 Community License
知识截止—
Benchmarks
评测成绩
| 评测基准 | 类别 | 分数 | 来源 |
|---|---|---|---|
| GSM-8K (CoT) | math reasoning | 95.1 | 来源 |
| ARC-C | reasoning general | 94.8 | 来源 |
| API-Bank | reasoning | 90.0 | 来源 |
| IFEval | general | 87.5 | 来源 |
| Multilingual MGSM (CoT) | math reasoning | 86.9 | 来源 |
| MMLU (CoT) | language reasoning math general | 86.0 | 来源 |
| MBPP ++ base version | reasoning general | 86.0 | 来源 |
| BFCL | general reasoning | 84.8 | 来源 |
| MMLU | general reasoning language math | 83.6 | 来源 |
| HumanEval | reasoning code | 80.5 | 来源 |
| DROP | reasoning math | 79.6 | 来源 |
| MATH (CoT) | math reasoning | 68.0 | 来源 |
| MMLU-Pro | language reasoning math general | 66.4 | 来源 |
| Multipl-E HumanEval | language general | 65.5 | 来源 |
| Multipl-E MBPP | general reasoning | 62.0 | 来源 |
| Nexus | general | 56.7 | 来源 |
| GPQA | reasoning general | 41.7 | 来源 |
| Gorilla Benchmark API Bench | reasoning code | 29.7 | 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 | $0.72 | $0.72 | 128K | — | — | ✗ | ✗ | ✗ |
价格单位：美元/百万 token，数据来自社区整理，仅供参考。