{
  "id": "moonshot-ai",
  "title": "Moonshot AI",
  "lang": "en",
  "source_lang": "en",
  "summary": "Moonshot AI is a Beijing-based artificial intelligence company founded in 2023 that develops the Kimi chatbot and a family of large language models, several released with open weights.",
  "category": "computer-science",
  "subcategory": "artificial-intelligence",
  "type": "organization",
  "aliases": [
    "Beijing Moonshot AI Technology Co., Ltd.",
    "Moonshot",
    "月之暗面",
    "Yuè Zhī Ànmiàn"
  ],
  "generated_at": "2026-10-16",
  "model": "pro/claude-haiku-5-5",
  "url": "https://aiwiki.page/en/moonshot-ai/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/moonshot-ai/",
      "json": "https://aiwiki.page/zh/moonshot-ai.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Moonshot AI (Chinese: 月之暗面) is a privately held artificial intelligence company headquartered in Beijing, China. It is best known for the Kimi chatbot and for the Kimi family of large language models, whose flagship releases are offered with open weights. The company was founded in March 2023 and, as of October 2026, is one of the most valuable independent AI developers in China."
    },
    {
      "heading": "Name",
      "text": "The Chinese name 月之暗面 means \"the dark side of the moon\". The company's English name follows the same idea. Its founders have described the metaphor as a way of framing the pursuit of artificial general intelligence as exploration of a poorly understood frontier. The chatbot name \"Kimi\" comes from the English given name that co-founder Yang Zhilin has used since his university years."
    },
    {
      "heading": "History",
      "text": "Moonshot AI was founded in March 2023 in Beijing by Yang Zhilin, Zhou Xinyu and Wu Yuxin. Yang had earlier studied at Tsinghua University, completed a machine learning PhD at Carnegie Mellon University, and worked as an intern at Google Brain and Meta AI. His research on Transformer-XL and XLNet, which addressed how language models retain long-range context, is often cited as the origin of Kimi's early emphasis on long inputs.\n\nThe company launched the Kimi chatbot in closed beta in October 2023 and opened it to the public on 16 November 2023. The first version supported a context of 128,000 tokens, which was unusually long at the time. In March 2024 it began testing a version with a context window of 2 million Chinese characters, and in July 2024 it introduced context caching in public beta.\n\nIn July 2025 Moonshot released Kimi K2, a trillion-parameter open-weight model. Later releases included K2 Thinking, K2.5, K2.6 (April 2026), a coding-oriented K2.7 variant (June 2026) and the flagship K3 (July 2026), which Moonshot describes as its largest open-weight model. Moonshot discontinued the hosted kimi-k2 API series on 25 May 2026 and directed users to newer models."
    },
    {
      "heading": "Models and technology",
      "text": "Kimi K2\n\nK2 uses a mixture-of-experts architecture with about one trillion total parameters, of which roughly 32 billion are active for each token. The base model was pre-trained on about 15.5 trillion tokens. Moonshot attributes the stability of training at this scale to its custom MuonClip optimizer, which builds on the Muon optimizer that the company has also described publicly. Moonshot has emphasised agentic use: K2 was post-trained on synthetic tool-use scenarios so that it could execute commands, call external tools and carry out multi-step tasks. Reported coding results on SWE-bench Verified are vendor figures, and their reproducibility depends on the test setup.\n\nLater models\n\nK2.6 supports a context window of about 262,000 tokens and is distributed in INT4 quantization, with a stated aim of lowering memory use and inference cost. Later versions add multimodal input and multi-agent \"swarm\" features. Kimi K3, released in July 2026, is reported to have about 2.8 trillion parameters and a context window of around one million tokens.\n\nProducts\n\nMoonshot's product line includes the Kimi chatbot and web application, Kimi Researcher, Kimi Agent, Kimi Code, Kimi Audio and the Kimina Prover model for formal mathematics. Its models are also offered through an API, which the company sells to businesses and developers. Commercial distribution has included partnerships with cloud providers such as Amazon Web Services.\n\nLicensing\n\nOpen-weight releases are published under a modified MIT license. The license permits downloading, running and fine-tuning the model, and requires very large deployments to credit \"Kimi K2\" in their interface. Training data and the full training recipe have not been released, and the Kimi app and API remain proprietary. Open-weight status therefore differs from fully open-source development in the usual sense."
    },
    {
      "heading": "Funding and corporate development",
      "text": "Moonshot's funding history has been reported by several outlets, and the figures differ across sources. According to Bloomberg, the company raised about $3.5 billion in a round closed on 29 July 2026 at a $35 billion valuation, a result that exceeded its initial target of $1 billion to $2 billion. The round was anchored by the National Artificial Intelligence Industry Investment Fund, a state-backed vehicle. Bloomberg reported in October 2026 that Moonshot had closed a final private round at a valuation of about $50 billion and was preparing an initial public offering in Hong Kong, targeted for the first quarter of 2027.\n\nEarlier rounds included a seed round of roughly $60 million in 2023, a round of more than $1 billion in February 2024 led by Alibaba at a valuation of about $2.5 billion, and a Series C in December 2025. Reported 2026 rounds include a $700 million raise in January at a $10 billion valuation and a round in May reported variously as $1 billion at $18 billion or $2 billion at $20 billion. Investors named in reports include Alibaba, Tencent, Meituan, HongShan (formerly Sequoia China), IDG Capital and the national fund. Secondary aggregators give differing totals and valuations, so the Bloomberg reports are the more reliable reference for the most recent rounds.\n\nMoonshot reported annual recurring revenue of about $200 million in April 2026, according to Wikipedia's infobox, and employed roughly 300 people in 2026. Its revenue remains small relative to its valuation, a gap that commentators have questioned."
    },
    {
      "heading": "Position in the industry",
      "text": "Moonshot is usually grouped with DeepSeek, Zhipu AI and the Qwen team of Alibaba among Chinese labs that compete at the frontier of large language models. Like DeepSeek and other Chinese developers, Moonshot has released advanced models openly, in contrast to the closed approach of several leading Western developers. Artificial Analysis rated K2 Thinking as the leading open-weight model on its intelligence index at the time of release, but also found that its high token consumption made it expensive to run."
    },
    {
      "heading": "Criticism and open questions",
      "text": "Benchmark comparisons between Kimi models and competing systems such as GPT-4.1, Claude and DeepSeek are largely reported by Moonshot itself. Analysts have noted that these comparisons use different metrics, sampling budgets and tool setups, so they do not by themselves establish overall leadership. Observers have also pointed out that heavy reliance on state-backed capital and on domestic demand may leave the company exposed to regulatory and geopolitical risk, particularly in relation to United States policy toward Chinese AI firms. Frequent model releases have also meant that some earlier APIs were retired within a short period, which affects developers who depend on specific versions."
    }
  ],
  "links": [
    "large-language-model",
    "artificial-intelligence",
    "carnegie-mellon-university",
    "google"
  ],
  "sources": [
    {
      "title": "Moonshot AI",
      "url": "https://en.wikipedia.org/wiki/Moonshot_AI"
    },
    {
      "title": "Kimi (AI)",
      "url": "https://en.wikipedia.org/wiki/Kimi_(AI)"
    },
    {
      "title": "Kimi (chatbot)",
      "url": "https://en.wikipedia.org/wiki/Kimi_(chatbot)"
    },
    {
      "title": "Moonshot AI Surpasses Funding Goal to Hit $35 Billion Value - Bloomberg",
      "url": "https://www.bloomberg.com/news/articles/2026-07-29/china-s-moonshot-ai-passes-funding-goal-to-hit-35-billion-value"
    },
    {
      "title": "Moonshot AI Said to Eye Early 2027 IPO After Value Hits $50 Billion - Bloomberg",
      "url": "https://www.bloomberg.com/news/articles/2026-10-06/moonshot-said-to-eye-early-2027-ipo-after-value-hits-50-billion"
    },
    {
      "title": "Moonshot AI’s $3.5 Billion Round Tripled Its Valuation in Six Months — and Beijing’s Fingerprints Are All Over It",
      "url": "https://finance.yahoo.com/technology/ai/articles/moonshot-ai-3-5-billion-072537342.html"
    },
    {
      "title": "Moonshot AI eyes US$30 billion valuation as China’s AI race intensifies",
      "url": "https://www.scmp.com/tech/article/3356348/moonshot-ai-eyes-us30-billion-valuation-chinas-ai-race-intensifies"
    },
    {
      "title": "How Much Did Moonshot AI Raise? Funding & Key Investors",
      "url": "https://www.clay.com/dossier/moonshot-ai-funding"
    },
    {
      "title": "Moonshot AI Revenue, Valuation, Funding & Investors",
      "url": "https://multiples.vc/private-comps/moonshot-ai"
    },
    {
      "title": "Moonshot AI Kimi",
      "url": "https://baike.baidu.com/en/item/Moonshot%20AI%20Kimi/4729433"
    },
    {
      "title": "What Is Kimi AI? A Complete Guide to Moonshot AI's Chatbot, Models, Pricing and Business Risks",
      "url": "https://www.simpliaxis.com/resources/what-is-kimi-ai"
    },
    {
      "title": "What Is Kimi? The Complete Guide to Moonshot AI's Chatbot and Models",
      "url": "https://emergent.sh/learn/what-is-kimi"
    },
    {
      "title": "Analysis of the Kimi K2 Open-Weight Language Model",
      "url": "https://intuitionlabs.ai/articles/kimi-k2-open-weight-llm-analysis"
    },
    {
      "title": "Kimi-K2 is the next open-weight AI milestone from China after Deepseek",
      "url": "https://the-decoder.com/kimi-k2-is-the-next-open-weight-model-breakthrough-from-china-after-deepseek/"
    },
    {
      "title": "Kimi K2: The Trillion-Parameter Open-Weight LLM",
      "url": "https://medium.com/@leucopsis/kimi-k2-the-trillion-parameter-open-weight-llm-9a656eb68cc5"
    },
    {
      "title": "What Is Kimi K2.6? Moonshot AI's Open-Weight Agent Model Explained - Verdent Guides",
      "url": "https://www.verdent.ai/guides/what-is-kimi-k2-6"
    },
    {
      "title": "What Is Kimi AI? Moonshot AI's Open-Weight Model, Explained",
      "url": "https://geotoolbox.ai/blog/what-is-kimi-ai"
    },
    {
      "title": "Every Kimi AI Model Explained and Compared (Sep, 2026) - Second Talent",
      "url": "https://www.secondtalent.com/resources/every-kimi-ai-model-explained-compared/"
    },
    {
      "title": "Moonshot AI valuation: $35B (Jul 2026) — funding & investors",
      "url": "https://valueaddvc.com/pulse/company/moonshot"
    }
  ],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}