跳到主要内容Skip to main content
投机取巧 Speculative Speculation

月之暗面（Moonshot AI）Moonshot AI　·　闭源谱系卡closed model lineage card　·　当前代current generation Kimi K3（API）Kimi K3 (API)（2026-06）

Kimi（闭源 API）Kimi (closed API)闭源模型 · 官方规格快照 · 带复核周期closed model · official spec snapshot · with review cycle

国产闭源旗舰，但它是本站唯一「API 与开放权重双轨」的家族：K3 权重在 Hugging Face / ModelScope 开放（自定义 Kimi K3 License），API 走 platform.kimi.com。选型时两条路径的许可与数据条款不同，分开读。The domestic closed flagship — yet the only family here with dual tracks: K3 weights are open on Hugging Face / ModelScope (custom Kimi K3 License) while the API runs on platform.kimi.com. Licences and data terms differ between the two paths — read them separately.

怎么用得到How you can use it：platform.kimi.com API 与 Kimi 应用；K3 权重同时开放（见本站开放权重侧的 kimi 谱系卡）——API 与开放权重是同一模型的两条使用路径。platform.kimi.com API and the Kimi app; K3 weights are also open (see the kimi family card on the open-weight side) — API and open weights are two usage paths to the same model.

开放权重侧本站收录 kimi-k3 系列（谱系卡 + 量化对比）；对话产品轻档案在 AI 工具站。On the open-weight side this site covers the kimi-k3 series (family card + quantisation); the chat light entry lives on the tools site.

中文生态China availability 国产 / 国内可直接用 / 备案状态 · 快照domestic / direct access / CAC filing · snapshot 2026-10-20

  • 出品属地Origin　国产domestic　出品方为月之暗面（北京）。Made by Moonshot AI (Beijing).
  • 国内下载Direct access　国内可直连direct OK　platform.kimi.com 大陆直连可达（2026-10-30 复测 200）。platform.kimi.com reachable direct from the mainland (measured 200 on 2026-10-30).
  • 备案状态CAC filing　已备案Filed：Beijing-MoonShot-20231016　公示名称listed as「Moonshot」　·　备案时间filed 2023-11-03　·　首批第 16 行（月之暗面科技）。Row 16 of the first batch (Moonshot AI).

备案信息以网信办《生成式人工智能服务已备案信息》公示名单为准（按批次滚动更新，本站只抄编号不解读）；「国内可直连」是本站从大陆网络直连实测的结果，不是厂商承诺。Filing status is quoted from the CAC public registry (updated batch by batch; this site copies the number and does not interpret it). "Direct access" is this site's own reachability measurement from a mainland network, not a vendor promise.

  • 网信办《生成式人工智能服务已备案信息》滚动公告页（13 个批次附件 PDF，2026-10-20 全量下载检索）：CAC · filed generative-AI service information (rolling announcement page, 13 batch attachment PDFs, full download & search on 2026-10-20): https://www.cac.gov.cn/2024-04/02/c_1713729983803145.htm

代次Generations 按官方发布/文档页记录as recorded by official releases and docs

  • Kimi K3Kimi K32026-06当前代current

    当前代：官方 API 文档 1M token 上下文（max_completion_tokens 默认 131072、最大 1048576）、始终启用思考（Preserved Thinking，推理强度 low/high/max 默认 max）、多模态；平台横幅「Kimi K3 旗舰模型已正式发布」。Current generation: official API docs list a 1M-token context (max_completion_tokens default 131072, max 1048576), always-on thinking (Preserved Thinking; reasoning effort low/high/max, default max), multimodal; the platform banner reads "Kimi K3 flagship officially released".

    官方页面 ↗official page ↗
  • Kimi K2.7 Code / K2.6Kimi K2.7 Code / K2.62026-04–06

    上一档：API 文档同页列 kimi-k2.7-code 与 kimi-k2.6 模型对象。Previous step: the same API docs page lists kimi-k2.7-code and kimi-k2.6 model objects.

    官方页面 ↗official page ↗

官方规格事实（每条挂出处）Official spec facts (each with a source) 截至核验日as of the verification date

  • 上下文窗口Context window　1M token（官方 API 文档：max_completion_tokens 最大可设 1048576）1M tokens (official API docs: max_completion_tokens can reach 1048576)出处 ↗source ↗
  • 思考模式Thinking mode　K3 始终启用思考 + Preserved Thinking；推理强度 low/high/max，默认 max（官方 API 文档）K3 always thinks, with Preserved Thinking; reasoning effort low/high/max, default max (official API docs)出处 ↗source ↗
  • 输入模态Input modalities　多模态理解与 Tool Calling（官方平台文档口径）Multimodal understanding and Tool Calling (official platform docs)出处 ↗source ↗

以上均为官方文档/发布页原文的转述，核验日之后厂商可能已改规格；本站不实测、不排名、不背书厂商自评，也不收录价格数字——定价请直接看官方定价页（出处节有链接）。The above transcribe official docs/release pages as of the verification date; vendors may have changed specs since. This site runs no benchmarks, ranks nothing, endorses no vendor self-assessments, and records no price figures — for pricing go to the official pricing page (linked in Sources).

出处Sources

  • Kimi 开放平台 API 文档（2026-10-30 直抓 200：K3 上下文/思考/推理强度参数）：Kimi open-platform API docs (direct fetch 200 on 2026-10-30: K3 context/thinking/reasoning-effort parameters): https://platform.kimi.com/docs/api/chat

核验快照：2026-10-30 · 复核周期 30 天，下次应在 2026-11-29 前复核。本站不实测、不排名，未知就写未知。Verification snapshot: 2026-10-30 · 30-day review cycle, next re-check due by 2026-11-29. No benchmarks, no rankings — unknown stays unknown.