Kimi K2.6vsKimi K2.5
在 12 个共同 benchmark 中，Kimi K2.6 整体领先：Kimi K2.6 领先 11 项，Kimi K2.5 领先 1 项，持平 0 项，平均分差 +6.27。
Kimi K2.611 项(92%)(8%)1 项Kimi K2.5
评测分数
按能力类目分组，每组内按分差大小排列；共 12 项。
Coding and Software Engineer
Kimi K2.6 领先 4/4| 评测项 | Kimi K2.6 | Kimi K2.5 | 分差 |
|---|---|---|---|
| SWE-Bench Pro - Public | 58.6013 / 54Thinking (With Tools) | 50.7041 / 54Thinking (With Tools) | +7.90 |
| LiveCodeBench | 89.607 / 123Thinking (No Tools) | 8516 / 123Thinking (No Tools) | +4.60 |
| SWE-bench Multilingual | 76.705 / 23Thinking (With Tools) | 7313 / 23Thinking (No Tools) | +3.70 |
| SWE-bench Verified | 80.2014 / 112Thinking (With Tools) | 76.8030 / 112Thinking (With Tools) | +3.40 |
General Knowledge
Kimi K2.6 领先 3/3| 评测项 | Kimi K2.6 | Kimi K2.5 | 分差 |
|---|---|---|---|
| HLE | 5415 / 172Thinking (With Tools + Internet) | 50.2027 / 172Thinking (With Tools) | +3.80 |
| LiveBench | 72.1728 / 115Thinking (No Tools) | 69.0742 / 115Thinking (No Tools) | +3.10 |
| GPQA Diamond | 90.5018 / 187Thinking (No Tools) | 87.6037 / 187Thinking (No Tools) | +2.90 |
Math and Reasoning
Kimi K2.6 领先 2/2| 评测项 | Kimi K2.6 | Kimi K2.5 | 分差 |
|---|---|---|---|
| IMO-AnswerBench | 868 / 21Thinking (No Tools) | 81.8016 / 21Thinking (No Tools) | +4.20 |
| AIME 2026 | 96.403 / 18Thinking (No Tools) | 92.5012 / 18Thinking (No Tools) | +3.90 |
AI Agent - Information Search
Kimi K2.6 领先 1/1| 评测项 | Kimi K2.6 | Kimi K2.5 | 分差 |
|---|---|---|---|
| BrowseComp | 83.2014 / 53Thinking (With Tools + Internet) | 60.6036 / 53Thinking (With Tools + Internet) | +22.60 |
AI Agent - Tool Usage
Kimi K2.6 领先 1/1| 评测项 | Kimi K2.6 | Kimi K2.5 | 分差 |
|---|---|---|---|
| Terminal Bench 2.0 | 66.7010 / 47Thinking (With Tools) | 50.8034 / 47Thinking (With Tools) | +15.90 |
Claw-style Agent Evaluation
Kimi K2.5 领先 1/1| 评测项 | Kimi K2.6 | Kimi K2.5 | 分差 |
|---|---|---|---|
| Claw Bench | 80.9019 / 29Thinking (With Tools) | 81.7018 / 29Thinking (With Tools) | -0.80 |
规格对比
| 字段 | Kimi K2.6 | Kimi K2.5 |
|---|---|---|
| 发布机构 | Moonshot AI | Moonshot AI |
| 发布时间 | 2026-04-20 | 2026-01-27 |
| 模型类型 | 推理大模型 | 多模态大模型 |
| 架构 | MoE 架构 | MoE 架构 |
| 参数规模 | 1万亿 | 1万亿 |
| 上下文长度 | 256K | 256K |
| 最大输出 | 暂无数据 | 16K |
API 调用价格
价格优先使用 DataLearner 配置的 API 记录；缺失项不做推测。
| 价格项 | Kimi K2.6 | Kimi K2.5 |
|---|---|---|
| 文本输入 | $0.95 / 1M tokens | $0.6 / 1M tokens |
| 文本输出 | $4 / 1M tokens | $3 / 1M tokens |
| 缓存读取 | $0.16 / 1M tokens | $0.1 / 1M tokens |
| 缓存写入 | $0.95 / 1M tokens | 暂无公开价格 |
小结
- Kimi K2.6在以下类目领先:Coding and Software Engineer (4/4)、General Knowledge (3/3)、Math and Reasoning (2/2)、AI Agent - Information Search (1/1)、AI Agent - Tool Usage (1/1)
- Kimi K2.5在以下类目领先:Claw-style Agent Evaluation (1/1)
12 个共同 benchmark 上，Kimi K2.6 平均高出 6.27 分。
单项差距最大的 benchmark：BrowseComp — Kimi K2.6 83.20，Kimi K2.5 60.60（分差 +22.60）。
本页正文由结构化模型、价格与 benchmark 数据生成，不使用实时 LLM 撰写。