Qwen3.8-Max-0902vsKimi K3
Qwen3.8-Max-0902 与 Kimi K3 在 2 个同模式 benchmark 中整体接近：Qwen3.8-Max-0902 领先 1 项，Kimi K3 领先 1 项，持平 0 项，平均分差 -3.06。另有 13 项测试模式不同，仅供参考。
Qwen3.8-Max-0902
阿里巴巴 · 2026-09-02 · 推理大模型
Kimi K3
Moonshot AI · 2026-07-16 · 推理大模型
Qwen3.8-Max-09021 项(50%)(50%)1 项Kimi K3
评测分数
按能力类目分组，每组内按分差大小排列；共 2 项。
仓库修复与多文件工程
Kimi K3 领先 1/1| 评测项 | Qwen3.8-Max-0902 | Kimi K3 | 分差 |
|---|---|---|---|
| FrontierSWE v2 | 17.8118 / 22Max (With Tools) | 25.8612 / 22Max (With Tools) | -8.06 |
跨能力综合测试
Qwen3.8-Max-0902 领先 1/1| 评测项 | Qwen3.8-Max-0902 | Kimi K3 | 分差 |
|---|---|---|---|
| SuperCLUE | 72.621 / 13Reported best (effort unspecified) | 70.685 / 13Reported best (effort unspecified) | +1.94 |
测试模式不同的成绩
共 13 项。两款模型的公开成绩来自不同测试模式，分数不直接可比，不计入胜负和平均分差。
| 评测项 | Qwen3.8-Max-0902 | Kimi K3 |
|---|---|---|
| Terminal-Bench 3.0 | 29Extra-High (With Tools) | 17.70Max (With Tools) |
| Program Bench | 28Extra-High (With Tools) | 17.50Max (With Tools) |
| FrontierMath Tier 4 v2 | 34.15Extra-High (With Tools) | 39.02Max (With Tools) |
| FrontierMath v2 | 65.61Extra-High (With Tools) | 72.18Max (With Tools) |
| MLS Bench | 50.10Extra-High (With Tools) | 48.30Max (With Tools) |
| AutomationBench | 50.80Extra-High (With Tools) | 46.70Max (With Tools) |
| DeepSWE | 69.30Extra-High (With Tools) | 68.51Max (With Tools) |
| NL2Repo-Bench | 64.90Extra-High (With Tools) | 58Max (With Tools) |
| SWE-Marathon | 44.80Extra-High (With Tools) | 42Max (With Tools) |
| Job Bench | 64Extra-High (With Tools) | 54.30Max (With Tools) |
| Toolathlon-Verified | 73.30Extra-High (With Tools) | 76.50Max (With Tools) |
| BabyVision | 93.80Extra-High (With Tools) | 85.70Max (With Tools) |
| MMMU-Pro | 82.70Extra-High (No Tools) | 83.40Max (With Tools) |
规格对比
| 字段 | Qwen3.8-Max-0902 | Kimi K3 |
|---|---|---|
| 发布机构 | 阿里巴巴 | Moonshot AI |
| 发布时间 | 2026-09-02 | 2026-07-16 |
| 模型类型 | 推理大模型 | 推理大模型 |
| 架构 | MoE 架构 | MoE 架构 |
| 参数规模 | 2.4万亿 | 2.8万亿 |
| 上下文长度 | 1M | 1M |
| 最大输出 | 131K | 1M |
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 | Qwen3.8-Max-0902 | Kimi K3 |
|---|---|---|
| 文本输入 | $2 / 1M tokens | ¥20 / 1M tokens |
| 文本输出 | $6 / 1M tokens | ¥100 / 1M tokens |
| 缓存读取 | $0.25 / 1M tokens | ¥2 / 1M tokens |
| 缓存写入 | $2.5 / 1M tokens | 暂无公开价格 |
小结
- Qwen3.8-Max-0902在以下类目领先:跨能力综合测试 (1/1)
- Kimi K3在以下类目领先:仓库修复与多文件工程 (1/1)
2 个同模式 benchmark 上，Kimi K3 平均高出 3.06 分。
单项差距最大的 benchmark：FrontierSWE v2 — Qwen3.8-Max-0902 17.81，Kimi K3 25.86（分差 -8.06）。
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。