API Hop

Kimi K3 API:价格与文本接入

查看 Kimi K3 的 Token 价格和 Chat Completions 示例,了解代码与文档分析的接入方式,并核对输入类型、推理行为和接口限制。

Kimi K3 API 价格与计费

官方价格

输入
$3.0000 /M
输出
$15.0000 /M
缓存输入
$0.3000 /M

API Hop 当前价格 降低 30%

输入
$2.1000 /M
输出
$10.5000 /M
缓存输入
$0.2100 /M

参考价格,尚未独立核验。比较成本前,请核对供应商当前价格、计费单位及适用条件。

Kimi K3 在线对话工具

模型设置

高级参数
0 个字符起(0/无固定上限)
无固定字符上限;完整请求仍受请求大小和模型上下文限制。此处不是 Token 计数。
0 – 2
0 – 1
-2 – 2
-2 – 2

对话

今天有什么可以帮你?

就绪

重置并开始新的对话?

这将清空当前对话记录和未发送的内容,并恢复默认参数。

API Hop 实测样例

2026 年 10 月 22 日经 API Hop 实际调用。每个结果仅代表所标模型与模式的一次请求,不代表成功率、速度或生产质量保证。以下模型回答保留原始语言。

带段落出处的发布决策提取

kimi-k3 · document-citations ·

原文节选

**Answer summary:** Production release is **not approved**. Staging established only export-formatting/export v7 verification — nothing about production. **Mei** owns the remaining duplicate-delivery tests; documentation ownership is **unassigned**.
完整模型回答(未经改写)
**Answer summary:** Production release is **not approved**. Staging established only export-formatting/export v7 verification — nothing about production. **Mei** owns the remaining duplicate-delivery tests; documentation ownership is **unassigned**.

| Conclusion | Exact supporting quote | Paragraph ID | Uncertainty |
|---|---|---|---|
| Production release is not approved; approval was withheld pending tests | "the release reviewer withheld production approval until duplicate-delivery tests pass" | P31 | Low — explicit statement |
| No approval or completed test recorded later (cross-check of P04's Friday shipping proposal, which "was not an approval") | "No completed test result or production approval is recorded in this document." | P36 | Low — explicit; P04 proposal superseded by P31/P36 |
| Staging established export formatting checks passed (repeated across components), and verified export v7 on synthetic data only | "staging verified export version 7 with three synthetic records. Production deployment and billing checks were not performed." | P18 | Low for staging result; production readiness not established |
| Staging did not measure production latency, charges, or customer outcomes | "no production latency, charge or customer outcome was measured" | P01 (and P02–P03, P05–P17, P19–P30, P32–P35, identical wording) | Low — explicit; applies to all component entries |
| Remaining duplicate-delivery tests are owned by Mei | "Mei owns those tests." | P31 | Low — explicit; tests scheduled, not completed ("Mei scheduled duplicate-delivery tests for October 16," P36) |
| Documentation has no owner; a documentation check was only planned | "Documentation ownership remains unassigned." | P31 (earlier: "Documentation ownership was still unassigned," P04) | Low that it is unassigned as of this document; future owner unknown. Note: component entries say "The named reviewer will check documentation next week" (e.g., P01), but no name is given and no completion is recorded |

**Unresolved questions:**
1. Did the duplicate-delivery tests scheduled for October 16 (P36) pass, and has the release reviewer since granted production approval?
2. Who will be assigned documentation ownership, and did the unnamed "named reviewer" ever complete the documentation check promised for "next week" (P01 et al.)?

输入文档为合成样本,不是最大上下文测试。结论与输入一致,但两处辅助引文标点未逐字保留:P36 的测试安排引文、P04 的文档归属引文均将原句末句号改为逗号,仍需人工核对引用。

完整请求与输入材料

POST /v1/chat-completions

{
  "model": "kimi-k3",
  "messages": [
    {
      "role": "user",
      "content": "Use only the following synthetic document. Answer: Is production release approved? What did staging establish? Who owns the remaining tests and documentation? Return a compact table of conclusion, exact supporting quote, paragraph ID and uncertainty, followed by two unresolved questions. Cross-check proposals against later updates; never infer a completed test or production success. Keep the answer under 500 words. This sample tests citation accuracy on this document, not maximum context length.\n\n[P01] Component 1 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P02] Component 2 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P03] Component 3 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P04] On October 10, the release coordinator proposed shipping export version 7 on Friday. The proposal was not an approval. Documentation ownership was still unassigned.\n\n[P05] Component 5 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P06] Component 6 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P07] Component 7 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P08] Component 8 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P09] Component 9 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P10] Component 10 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P11] Component 11 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P12] Component 12 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P13] Component 13 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P14] Component 14 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P15] Component 15 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P16] Component 16 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P17] Component 17 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P18] On October 11, staging verified export version 7 with three synthetic records. Production deployment and billing checks were not performed.\n\n[P19] Component 19 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P20] Component 20 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P21] Component 21 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P22] Component 22 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P23] Component 23 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P24] Component 24 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P25] Component 25 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P26] Component 26 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P27] Component 27 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P28] Component 28 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P29] Component 29 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P30] Component 30 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P31] On October 12, the release reviewer withheld production approval until duplicate-delivery tests pass. Mei owns those tests. Documentation ownership remains unassigned.\n\n[P32] Component 32 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P33] Component 33 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P34] Component 34 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P35] Component 35 review: staging checks passed for export formatting. Production rollout has not been approved by this entry. The named reviewer will check documentation next week; no production latency, charge or customer outcome was measured. This paragraph records a proposed review, not completion.\n\n[P36] On October 13, Mei scheduled duplicate-delivery tests for October 16. No completed test result or production approval is recorded in this document."
    }
  ],
  "stream": false,
  "max_tokens": 4096
}
响应元数据与返回用量
{
  "requestId": "339d0045e51385a03300d75fa74f0863",
  "jobId": null,
  "responseModel": "kimi-k3",
  "responseAt": "2026-10-22T12:28:31.361Z",
  "finishedAt": "2026-10-22T12:28:31.364Z",
  "usage": {
    "prompt_tokens": 2058,
    "completion_tokens": 1086,
    "total_tokens": 3144,
    "prompt_tokens_details": {
      "cached_tokens": 0,
      "cache_write_tokens": 0,
      "text_tokens": 0,
      "audio_tokens": 0,
      "image_tokens": 0
    },
    "completion_tokens_details": {
      "text_tokens": 0,
      "image_tokens": 0,
      "audio_tokens": 0,
      "reasoning_tokens": 0
    },
    "input_tokens": 0,
    "output_tokens": 0,
    "input_tokens_details": null,
    "output_tokens_details": null,
    "claude_cache_creation_5_m_tokens": 0,
    "claude_cache_creation_1_h_tokens": 0
  },
  "outputSha256": "e3ba3715b7facfd224f3e1982c457c3647c33114774768190023cef8b1eef396",
  "responseEvidence": {
    "sha256": "2f1507aa711a70a17163506c7420517b599e4564fcf69aaef6db8cf5d7b09506",
    "bytes": 6024,
    "phase": "submission",
    "http": 200,
    "requestId": "339d0045e51385a03300d75fa74f0863",
    "observedAt": "2026-10-22T12:28:31.361Z"
  },
  "assets": []
}
下载公开证据记录

Kimi K3 提示词示例

这些是 Kimi K3 的评估提示词,不是模型实测输出。仅分析提供的文本,不执行代码、不检查生产系统,也不证明长上下文表现。请核对引文与验收标准,并在用于生产前检查响应元数据和用量。

仅使用这份文档样本:[P1] 导出修复通过了预发布检查。[P2] 有人提议周五发布,但未获批准。[P3] Mei 将在周二审核错误提示。[P4] 文档工作尚未分配负责人。以表格返回结论、准确引文、段落编号和不确定性。生产发布是否已批准?谁负责文档?缺失事实标为未知。每段引文必须与样本一致,不要由预发布成功推断生产成功。这个小样本不测试最大上下文长度。

将这些合成记录提取为含 id、owner、deadline、evidence 字段的 JSON:[R1] 问题 41:Ana 负责导出修复;尚未约定期限。[R2] 问题 42:审核期限为 10 月 15 日;未指定负责人,未说明年份。缺失值使用 null,保留原文的不完整日期。evidence 引用原文,不要猜测年份或负责人。逐字段核对来源;这只是文本整理,不是数据库操作。

一个虚构的两人团队每天运行 2000 份报告,允许在 30 分钟内完成。已有数据库,没有队列服务;任务须在重启后恢复并避免重复发送。比较数据库批处理与独立队列,给出决策表、有条件的建议、三个未知项及一个会改变选择的条件。不要编造基准、成本或已经验证的保证。

什么是 Kimi K3

面向长流程编程与知识工作的原生多模态模型

我们将 Kimi K3 定位为面向复杂任务的推理与 Agent 模型,重点关注需要持续理解上下文、调用工具并根据结果修正方案的工作场景。它由 Moonshot AI 开发,是多模态模型,具备原生视觉能力,可以理解文本、图片和视频,适用方向包括项目编程、资料研究和多步骤知识工作。

它的多模态能力能够扩展它的应用场景,比如图片理解/OCR,它和传统OCR方案相比的优点在于,你不需要单独训练模型,就能够有比较好的理解能力。

支持百万 Token 上下文的开放权重 MoE 模型

Kimi K3 采用混合专家架构,也就是 MoE,总参数为 2.8 万亿,每个 Token 激活约 1040 亿参数,模型上下文窗口为 1,048,576 Token。官方已经开放完整模型权重,采用 Kimi K3 License。

我们认为,百万 Token 上下文的实际价值,在于可以把更多相关资料放进同一次分析,例如项目规范、多个模块的代码和历史讨论。对于有自主部署需求的团队,开放权重也给部分企业和用户提供了自部署的场景。

始终开启思考,支持三档推理强度

Kimi K3 始终进行思考。当前 Kimi 官方 API 通过 reasoning_effort 提供 low、high、max 三档推理强度,默认使用 max。

我们建议根据任务难度安排推理预算:简单分类和资料提取可以先使用 low,多步骤分析可以从 high 开始,需要反复检查约束的复杂任务再比较 max。我们从其他从模型的测试经验来看,max的综合收益并不一定是最高的,所以要结合场景来做测试和判断。

RequestBeam 的 Kimi K3 在线工具与 API

我们希望通过 RequestBeam,让你更快地了解 Kimi K3 在自己任务中的真实表现。你只需要登录、创建 API Key,并完成少量充值,就可以通过在线工具开始体验,无需先编写 API 对接代码,也不必为了短期评估购买固定套餐。

Kimi K3 的核心功能与领域优势

长周期软件开发:从理解需求到修复问题、运行测试

我们认为,长周期编程是 Kimi K3 最值得优先测试的能力。它可以在 Agent 框架提供文件、终端和测试工具的情况下,连续完成代码阅读、方案修改、测试执行和问题修复,而不只是根据一句需求生成一段代码。

在我们收集的客户的使用分享中,有用户提到,Kimi K3 帮助他继续推进了此前停滞的项目,并完成多个连续开发阶段。也有开发者表示,它在跨文件分析和寻找模块关系方面表现不错,但第一次提出的修改方案不一定安全,有时还会引入新的问题。

我们觉得,这恰好说明了 Kimi K3 在真实开发中的使用方式:不要只让模型“帮我们修复这个 BUG”,而是提供明确的测试用例、不能改变的既有行为,以及允许修改的文件范围。

对于后端重构、复杂数据处理和生产环境代码,我们建议让它先解释修改计划,再执行修改,最后通过自动化测试确认结果。即使模型已经报告任务完成,也不应跳过代码审查。

可视化开发:结合界面截图持续优化前端与交互效果

Kimi K3 的视觉理解能力,使它能够将界面截图、设计要求和代码放在同一任务中处理。

例如,我们可以先让模型根据参考图片生成网页,再通过浏览器工具获取实际渲染截图,让模型比较页面与设计稿之间的差异,继续调整布局、颜色、字体和交互细节。

在我们的客户群反馈中,有用户对 Kimi K3 的前端、可视化项目和小游戏制作给出了积极评价。

我们认为,Kimi K3 比较适合产品原型、活动页面、可视化仪表盘和需要反复对照截图的开发任务。尤其是独立开发者,可以先用它生成可运行的版本,再逐步完善页面体验。

Agent 工具协作:执行多步骤任务,需要给于明确的边界

Kimi K3 可以配合外部工具进行文件读写、终端操作、资料检索以及多步骤自动化工作。模型负责判断下一步行动,实际工具调用和执行环境则由应用程序或 Agent 框架提供。

我们特别关注一个现象:同一个模型放在不同 Agent 框架中,最终任务完成率可能有明显差异。

此外,Moonshot AI 官方也提醒,Kimi K3 可能在用户意图不明确时主动做出超出预期的决定。因此,在需要严格控制行为的任务中,我们建议通过系统提示词或 AGENTS.md 明确操作权限。

Kimi K3 基准测试与性能表现

我们在这一部分分别使用 Moonshot 的发布期官方测试,以及截至 2026 年 10 月 7 日可查的 Artificial Analysis 独立评测。下述官方 Kimi K3 成绩均采用 max 思考强度;我们同时保留基准版本和关键测试条件,方便你判断它们与实际任务的关系。

软件工程:DeepSWE 67.5%,ProgramBench 77.8%

Moonshot 公布的 Kimi K3 成绩中,DeepSWE v1.1 为 67.5%,ProgramBench 为 77.8%,这两项 K3 测试使用 Kimi Code 框架。我们把它们作为模型处理代码仓库任务与持续编程工作的参考。

对于准备接入编程助手的团队,我们建议分别测试已有问题修复、小功能开发和小型项目实现。评估重点可以放在一次完成率、失败后的修正能力,以及最终代码需要多少人工返工。这样的结果更容易对应到开发成本。

终端任务:Terminal-Bench 2.1 与 4.0 需要分别解读

官方发布期测试中,Kimi K3 配合 Kimi Code,在 Terminal-Bench 2.1 上取得 88.3%。当前 Artificial Analysis 的 Terminal-Bench 4.0 测试中,K3 Max 的成绩约为 13%。

两组结果对应不同任务集和执行框架。AA 的 4.0 测试包含 66 个更难的终端任务,使用 mini-swe-agent,并对每个任务重复三次。在实际业务场景中,我们建议你使用其他harness进行测试。

研究与工具调用:BrowseComp 91.2%,MCPMark-Verified 94.5%

官方报告中,Kimi K3 的 BrowseComp 成绩为 91.2%,MCPMark-Verified 为 94.5%。其中 BrowseComp 的 91.2% 使用了在 300K Token 时触发上下文压缩的策略;直接使用 1M 上下文且不进行上下文管理时,成绩为 90.4%。

我们认为,这组结果说明,研究和工具协这块是 Kimi K3 的强项,同时也提醒我们关注应用如何管理上下文,从我们的经验来看,上下文管理对于实际效果的影响会很大。

长文推理:AA-LCR v1.1 达到 88.7%

Artificial Analysis 的 AA-LCR v1.1 专项页面记录,Kimi K3 Max 的成绩为 88.7%。这项评测包含 100 道需要综合长文档信息的问题,涉及公司报告、论文和行业资料等内容;文档长度约为 1 万至 10 万 Token。

我们将这项结果视为 Kimi K3 跨资料分析能力较有说服力的证据。对于接近百万 Token 的实际输入,我们仍建议单独测试,按照我们以往的测试经验来看,如果输入贴近模型允许的上限,它的输出并不会特别好。

综合指数与响应速度:44 分,输出约 45 Token/秒

Artificial Analysis Intelligence Index v4.3.2 将 Kimi K3 Max 记为 44 分。其输出速度约为每秒 45 Token,首个 Token 约需 3.92 秒,计入思考后首个答案 Token 约需 48.58 秒。

我们会分别检测开始返回数据、出现正式答案和整个任务完成的时间。对于交互频繁的功能,这些时间会影响用户的使用体验,对于复杂分析,则需要结合业务场景的结果质量判断。

Kimi K3 与 DeepSeek V4 Pro、V4.1 Flash、Claude Fable 5.1、GPT-5.6 Sol 同类模型对比

我们使用 Artificial Analysis v4.3.2 的同版本数据进行比较。以下均为 Max 档位;Claude Fable 5.1 保留评测页标注的 Default Fallback 配置。AA-LCR 在比较页中以整数百分比显示,因此 Kimi K3 的 88.7% 显示为 89%。

表中的任务成本是 AA 按评测权重计算的平均成本,包含输入、缓存、思考和答案等用量,供模型选型参考;RequestBeam 的实际费用按平台价格与请求用量计算。成本口径说明

模型与测试档位综合指数AA-LCR v1.1加权单任务成本
Kimi K3 Max4489%$2.00
DeepSeek V4 Pro 0813 Max3680%$0.67
DeepSeek V4.1 Flash Max3984%$0.27
Claude Fable 5.1 Max,Default Fallback5385%$7.63
GPT-5.6 Sol Max4784%$1.99
Artificial Analysis v4.3.2 2026-10-07 · 全部为 Max 档位
综合指数 分 · 越高越好
  • Kimi K3 Max 44
  • DeepSeek V4 Pro 0813 Max 36
  • DeepSeek V4.1 Flash Max 39
  • Claude Fable 5.1 Max默认后备配置 53
  • GPT-5.6 Sol Max 47
AA-LCR v1.1 百分比 · 越高越好
  • Kimi K3 Max 89%
  • DeepSeek V4 Pro 0813 Max 80%
  • DeepSeek V4.1 Flash Max 84%
  • Claude Fable 5.1 Max默认后备配置 85%
  • GPT-5.6 Sol Max 84%
加权单任务成本 美元/任务 · 越低越好
  • Kimi K3 Max $2.00
  • DeepSeek V4 Pro 0813 Max $0.67
  • DeepSeek V4.1 Flash Max $0.27
  • Claude Fable 5.1 Max默认后备配置 $7.63
  • GPT-5.6 Sol Max $1.99
AA-LCR 按整数显示:Kimi K3 88.7% → 89%。 AA 任务成本包含输入、缓存、思考和答案等用量,不是 API Hop 平台价格。

数据来源:V4 Pro 对比、V4.1 Flash 对比、Fable 5.1 对比、GPT-5.6 Sol 对比。

对比 DeepSeek V4 Pro:长文分析表现更强,评测任务成本更高

在这组评测中,Kimi K3 的综合指数、长文推理与知识工作成绩高于 DeepSeek V4 Pro 0813。我们会优先把 K3 用于资料之间联系较复杂、需要综合多种证据的任务,并检查这种能力差异是否能减少人工整理和复核的时间。AA 同口径对比

对于大量重复、验收标准清楚的文本任务,我们也会测试 V4 Pro。它在这套评测中的任务成本更低;如果输出已经达到业务要求,成本差异就可能成为扩大使用规模时的重要因素。

对比 DeepSeek V4.1 Flash:综合与长文占优,自动化任务另有取舍

Kimi K3 在综合指数与 AA-LCR 上领先,但在 AutomationBench-AA 中,V4.1 Flash 为 69%,K3 为 58%;在 Terminal-Bench 4.0 中,两者分别为 27% 和 13%。我们认为,这说明不同任务类型会改变模型排序。AA 同口径对比

如果应用主要处理长资料分析,我们建议重点比较 K3;如果任务集中在流程自动化、终端操作和高频调用,我们会把 V4.1 Flash 列为重要候选,并同时比较成功率与完整执行成本。

对比 Claude Fable 5.1:终端编程存在差距,长文任务值得单独比较

在上述 Default Fallback 配置下,Claude Fable 5.1 的综合指数为 53,Kimi K3 为 44;Terminal-Bench 4.0 分别为 52% 和 13%。我们会把 Fable 5.1 作为复杂终端编程的重要对照。K3 则在 AA-LCR v1.1 上取得更高的成绩,并具有更低的评测任务成本。AA 同口径对比

对于阅读多份资料后回答明确问题的应用,我们建议单独测试 K3 的效果。对于需要持续操作终端、修复环境和完成工程验证的任务,我们会更关注首次成功率与人工介入次数,再判断额外模型费用是否值得。

对比 GPT-5.6 Sol:长文推理各有表现,完整任务成本接近

Kimi K3 的 AA-LCR 得分高于 GPT-5.6 Sol;GPT-5.6 Sol 的综合指数和 Terminal-Bench 4.0 成绩更高。我们尤其关注两者的任务成本:在这套评测中分别约为 2.00 美元和 1.99 美元,已经十分接近。AA 同口径对比

我们因此建议把完整任务跑完以后再比较费用,包括思考、工具交互、重试和最终答案。对于资料密集的分析工作,可以优先比较 K3;对于终端工程任务,可以把 GPT-5.6 Sol 作为重点对照,最后根据业务中的实际完成情况决定路由。

为什么通过 API Hop 使用 Kimi K3?

按需计费与模型成本优化

按当前 API Hop 单价和实际用量计费,按需充值。

统一 AI模型 与数据 API,一站式多模型接入

一个密钥多个AI模型和数据模型,节省对接成本,一键切换,业务需要的场景必然是AI+数据

调用记录与应用权限控制,每一次调用明明白白

每一笔调用记录都可以在后台查到,可以设置多个密钥对应不同的应用

按任务验收模型输出

通过 API Hop 使用文档列出的模型 ID 与支持参数。按自己的验收标准评估输出,投入生产前核对用量记录与响应元数据;模型名称或演示样例不保证每次结果。

如何在 API Hop 中使用 Kimi K3

在后台创建完密钥后,就可以根据文档来接入 Kimi K3。

准备账户并进行在线测试

登录并确认余额,创建一个 API 密钥。

在 Playground 中体验

可以在 Playground 中体验模型的能力,确定最佳的模型使用方案。

将 Kimi K3 集成到你的业务中

根据我们提供的开发文档,将 Kimi K3 集成到业务中。

后续的模型升级

升级前先确认 API Key 的模型列表已返回新型号的准确 ID。仅在新旧模型支持相同接口和请求字段时更换 model,并验证返回格式、流式行为、输出上限与费用后再切换生产流量。涉及异步图片或视频任务时,还应验证任务提交、状态查询与结果获取流程。

Kimi K3 API 常见问题

通过 RequestBeam 调用 Kimi K3,接口地址和模型名称怎么填写?

我们目前提供的 Kimi K3 接入地址是 POST https://api.requestbeam.com/v1/chat-completions,模型标识填写 kimi-k3,认证使用 RequestBeam API Key。可以先按下面的最小请求确认接入,再逐项增加可选参数。RequestBeam Kimi API 文档

curl -X POST "https://api.requestbeam.com/v1/chat-completions" \
  -H "Authorization: Bearer $REQUESTBEAM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3",
    "messages": [
      {"role": "user", "content": "请将这项需求整理为目标、约束和验收条件:为内部员工创建一个支持邮箱登录和密码重置的管理后台。"}
    ],
    "stream": false
  }'

此 API Hop 模型使用统一 Chat Completions 接口,而不是 OpenAI Responses 或 Claude Messages。

Kimi K3 支持流式输出和连续追问吗?

我们提供流式和非流式两种响应方式。设置 stream: true 时返回 SSE,设置 stream: false 或省略该字段时返回完整 JSON。对于需要逐步展示结果的界面,我们建议使用流式响应。RequestBeam Kimi API 文档

连续追问需要应用按接入协议维护对话历史,并在后续请求中携带相关上下文。在线工具中的 Memory 用来加入当前对话中可复用的历史轮次;我们建议根据问题保留必要资料,并关注历史增加带来的输入 Token 成本。在线工具的上下文说明

Playground 记忆不会跨会话长期保留。关闭后历史仍可见,但不会随新消息发送。失败、停止、截断或不可复用的回复不会作为正常历史继续附带。

0

Kimi K3 支持百万 Token 上下文,为什么回答仍可能被截断?

我们会先分别检查上下文容量和输出预算。RequestBeam 的 max_tokens 用于配置输出额度,思考与最终答案共享这部分预算。额度太小时,即使还没有输出完整正文,也可能以 finish_reason: "length" 结束。RequestBeam 输出预算说明

遇到 length 时,我们建议检查是否给思考和答案留出了足够空间,再决定增加预算或分步完成任务。如果报错指向上下文超限,则需要精简无关历史和重复资料。

最后我们需要澄清一个观点:百万 token 上下文,不代表它真的能在百万 token 级别返回好的结果,如果你追求结果的正确性,那么控制上下文的长度目前为止是你唯一的选择。

API 客户端应区分 reasoning_content 与最终回答的 content。

0

调用 Kimi K3 时,400、401、402 和 429 分别怎么处理?

我们建议先检查 HTTP 状态,再查看响应中的数字 code。RequestBeam 的 Chat Completions 文档将常见错误定义如下。RequestBeam 错误响应文档

HTTP 状态数字 code含义我们建议的处理方式
40010000请求无效检查 model、messages、字段类型与可选参数
40120000认证缺失或无效检查 RequestBeam API Key 和 Bearer 认证头
40240000可用余额不足检查账户余额及本次请求需要的额度
42940001请求触发限流降低并发和频率,并采用退避重试

我们建议从最小请求开始排查,再逐项恢复参数。HTTP 状态与响应中的 code 需要分别处理;如果问题持续出现,可以保留响应中的请求标识和相关参数,方便我们定位具体环节。

403 应检查当前 Key 可用模型目录中的访问权限。502、503、504 和 SSE 流中错误应根据完整错误响应排查。

Kimi K3 Playground 可以上传图片、PDF 或视频吗?

本页提供文本对话,没有文件或图片上传入口。K3 原生具备视觉理解能力,但需要支持对应输入的接口;这里不会自动下载链接、读取外部文件或联网搜索。可先提取相关文本,再提交分析。

一次请求可以返回多个方案或保证 JSON 正确吗?

Kimi K3 在 API Hop 中仅支持 n=1。可以要求在一个回答里列出多个方案。需要 JSON 时说明字段、类型与缺失值处理,并在应用中做语法、Schema 和业务校验。

相关文本 API