Kimi K3vsGPT-5.6 Sol
在 21 个共同 benchmark 中，GPT-5.6 Sol 整体领先：Kimi K3 领先 5 项，GPT-5.6 Sol 领先 16 项，持平 0 项，平均分差 +0.51。
Kimi K3
Moonshot AI · 2026-07-16 · 推理大模型
GPT-5.6 Sol
OpenAI · 2026-06-26 · 推理大模型
Kimi K35 项(24%)(76%)16 项GPT-5.6 Sol
评测分数
按能力类目分组，每组内按分差大小排列；共 21 项。
Productivity Knowledge
Kimi K3 领先 3/5| 评测项 | Kimi K3 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Harvey Lab-AA | 94.601 / 6Max (With Tools) | 2.506 / 6Max (With Tools) | +92.10 |
| GDPval-AA v2 | 1,68610 / 27Max (With Tools) | 1,7289 / 27Max (With Tools) | -42 |
| AA-Briefcase | 1,5288 / 20Max (With Tools) | 1,4949 / 20Max (With Tools) | +34.28 |
| AA-AnalystAgent | 38.757 / 12Max (With Tools + Internet) | 47.505 / 12Max (With Tools + Internet) | -8.75 |
| AutomationBench | 46.707 / 17Max (With Tools) | 45.808 / 17Max (With Tools) | +0.90 |
AI Agent - Tool Usage
GPT-5.6 Sol 领先 4/4| 评测项 | Kimi K3 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Terminal-Bench 4.0 | 12.6013 / 16Max (With Tools) | 39.905 / 16Max (With Tools) | -27.30 |
| Terminal-Bench 3.0 | 17.708 / 11Max (With Tools) | 34.402 / 11Max (With Tools) | -16.70 |
| Terminal-Bench-Science 0.1 | 7.109 / 11Max (With Tools) | 22.404 / 11Max (With Tools) | -15.30 |
| CyberGym | 805 / 8Max (With Tools) | 84.502 / 8Max (With Tools) | -4.50 |
Coding and Software Engineer
GPT-5.6 Sol 领先 2/3| 评测项 | Kimi K3 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| DeepSWE | 67.5011 / 38Max (With Tools) | 736 / 38Max (With Tools) | -5.50 |
| Program Bench | 17.5010 / 11Max (With Tools) | 237 / 11Max (With Tools) | -5.50 |
| NL2Repo-Bench | 586 / 16Max (With Tools) | 56.809 / 16Max (With Tools) | +1.20 |
Multimodal Understanding
GPT-5.6 Sol 领先 3/3| 评测项 | Kimi K3 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| ZeroBench Main | 414 / 6Max (With Tools) | 531 / 6Max (With Tools) | -12 |
| Chartography | 68.106 / 8Max (With Tools) | 79.903 / 8Max (With Tools) | -11.80 |
| BabyVision | 85.705 / 9Max (With Tools) | 88.904 / 9Max (With Tools) | -3.20 |
Agent Level Benchmark
GPT-5.6 Sol 领先 2/2| 评测项 | Kimi K3 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| APEX-Agents | 415 / 6Max (With Tools) | 56.703 / 6Max (With Tools) | -15.70 |
| τ³-Banking | 33.406 / 13Max (With Tools) | 44.331 / 13Max (With Tools) | -10.93 |
Long Context
GPT-5.6 Sol 领先 1/1| 评测项 | Kimi K3 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| AA-LCR | 74.708 / 29Max (No Tools) | 77.674 / 29Max (No Tools) | -2.97 |
Math and Reasoning
GPT-5.6 Sol 领先 1/1| 评测项 | Kimi K3 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| FrontierMath v2 | 72.1813 / 58Max (No Tools) | 89.121 / 58Max (No Tools) | -16.94 |
Text Embedding
GPT-5.6 Sol 领先 1/1| 评测项 | Kimi K3 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Context Arena | 71.7559 / 126Max (No Tools) | 97.632 / 126Max (No Tools) | -25.88 |
Writing and Creative Capabilities
Kimi K3 领先 1/1| 评测项 | Kimi K3 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Creative Writing | 2,0714 / 106Normal (No Tools) | 1,9636 / 106Normal (No Tools) | +107.20 |
规格对比
| 字段 | Kimi K3 | GPT-5.6 Sol |
|---|---|---|
| 发布机构 | Moonshot AI | OpenAI |
| 发布时间 | 2026-07-16 | 2026-06-26 |
| 模型类型 | 推理大模型 | 推理大模型 |
| 架构 | MoE 架构 | 稠密模型 |
| 参数规模 | 2.8万亿 | 暂无数据 |
| 上下文长度 | 1M | 1.05M |
| 最大输出 | 1M | 128K |
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 K3 | GPT-5.6 Sol |
|---|---|---|
| 文本输入 | ¥20 / 1M tokens | $4 / 1M tokens |
| 文本输出 | ¥100 / 1M tokens | $20 / 1M tokens |
| 缓存读取 | ¥2 / 1M tokens | $0.4 / 1M tokens |
| 缓存写入 | 暂无公开价格 | $5 / 1M tokens |
小结
- Kimi K3在以下类目领先:Productivity Knowledge (3/5)、Writing and Creative Capabilities (1/1)
- GPT-5.6 Sol在以下类目领先:AI Agent - Tool Usage (4/4)、Coding and Software Engineer (2/3)、Multimodal Understanding (3/3)、Agent Level Benchmark (2/2)、Long Context (1/1)、Math and Reasoning (1/1)、Text Embedding (1/1)
21 个共同 benchmark 上，Kimi K3 平均高出 0.51 分。
单项差距最大的 benchmark：Creative Writing — Kimi K3 2,071，GPT-5.6 Sol 1,963（分差 +107.20）。
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。