Kimi K2 base model is a state-of-the-art mixture-of-experts (MoE) language model with 32 billion activated parameters and 1 trillion total parameters. Trained on 15.5 trillion tokens with the MuonClip optimizer, this is the foundation model before instruction tuning. It demonstrates strong performance on knowledge, reasoning, and coding benchmarks while being optimized for agentic capabilities.
发布日期2025年7月11日
参数规模1.0T
上下文长度—
许可证MIT
知识截止—
Benchmarks
评测成绩
| 评测基准 | 类别 | 分数 | 来源 |
|---|---|---|---|
| C-Eval | general reasoning | 92.5 | 来源 |
| GSM8k | math reasoning | 92.1 | 来源 |
| MMLU-redux-2.0 | language reasoning math general | 90.2 | 来源 |
| MMLU | general reasoning language math | 87.8 | 来源 |
| TriviaQA | general reasoning | 85.1 | 来源 |
| EvalPlus | reasoning code | 80.3 | 来源 |
| CSimpleQA | general language | 77.6 | 来源 |
| MATH | math reasoning | 70.2 | 来源 |
| MMLU-Pro | language reasoning math general | 69.2 | 来源 |
| GPQA | reasoning general | 48.1 | 来源 |
| SuperGPQA | reasoning general math legal healthcare finance chemistry economics physics | 44.7 | 来源 |
| SimpleQA | general reasoning | 35.3 | 来源 |
| LiveCodeBench v6 | reasoning general | 26.3 | 来源 |
Pricing
API 价格对比
暂无 API 价格。